Grazing by two species of limpets on artificial reefs in the northwest Mediterranean

Transcription

1 Journal of Experimental Marine Biology and Ecology 255 (2000) locate/ jembe Grazing by two species of limpets on artificial reefs in the northwest Mediterranean * Fabio Bulleri, Massimo Menconi, Francesco Cinelli, Lisandro Benedetti-Cecchi Dipartimento di Scienze dell Uomo e dell Ambiente, via A. Volta 6, Pisa, Italy Received 14 January 2000; received in revised form 7 July 2000; accepted 1 August 2000 Abstract The exteive presence of artificial reefs in marine coastal habitats demands a better understanding of the extent to which these structures can be coidered surrogates of natural rocky shores for populatio of plants and animals. The primary aim of this study was to test the hypothesis that removing limpets from the midlittoral of artificial breakwaters in the northwest Mediterranean led to changes in assemblages similar to those observed on rocky shores in the same area. Orthogonal combinatio of the presence/ absence of two species of limpets, P. aspera and P rustica, were produced using manual removals from June 1997 to February To test the hypothesis that the effects of limpets were variable at spatial scales comparable to those investigated on rocky shores, we repeated the experiment at two locatio te of kilometres apart, and on two reefs within each location a few kilometres apart. The results revealed strong and relatively coistent negative effects of limpets on filamentous algae, whereas interactio with other members of assemblages were complex and variable. Several taxa (Cyanophyta, encrusting and articulated coralline algae, Ralfsia and Rissoella) were abundant at one location but nearly absent at the other. This large-scale variability in patter of distribution generated incoistencies in the effects of limpets between locatio. Within locatio, several effects of P. aspera and P. rustica were observed, ranging from independent effects on some organisms, to additive or interactive effects on others. Apparently, the removal of filamentous algae by limpets resulted in positive indirect effects on Ralfsia and Rissoella. Collectively, these effects were comparable to those described for rocky shores in the northwest Mediterranean. The processes accounting for large-scale variation in grazing, however, appeared different between the natural and the artificial habitat Elsevier Science B.V. All rights reserved. *Corresponding author. Present address: Centre for Research on Ecological Impacts of Coastal Cities, Marine Ecology Laboratories A 11, University of Sydney, Sydney NSW 2006, Australia. Tel.: ; fax: address: fbulleri@bio.usyd.edu.au (F. Bulleri) / 00/ $ see front matter 2000 Elsevier Science B.V. All rights reserved. PII: S (00)

2 2 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Keywords: Artificial reefs; Intertidal; Limpets; Patella aspera; Patella rustica; Grazing; Indirect effects; Spatial heterogeneity 1. Introduction Artificial reefs and breakwaters are common structures in marine coastal habitats. Breakwaters are used to build marinas and harbours and to protect sandy shores from erosion. Subtidal reefs are thought to attract species previously absent in the area, and are often used to restore over-fished populatio by increasing the complexity of the habitat and the availability of shelter (Seaman et al., 1989; Carr and Hixon, 1997). Since these structures act as surrogates of rocky shores, it is important to understand whether they support assemblages that are comparable to those found on natural substrata and if the ecological processes operating in the two environments are also similar. Most of the ecological studies on artificial reefs and breakwaters have been confined to subtidal habitats. There have been studies assessing the effectiveness of artificial reefs in mitigating losses of commercial species due to human disturbance (Ambrose, 1994; Carr and Hixon, 1997), some experimental investigatio on the effects of dispersal, recruitment, physical processes and biological interactio in influencing the structure of epibiota on artificial structures (Grosberg, 1982; Keough and Butler, 1983; Keough, 1984; Breitburg, 1985; Anderson and Underwood, 1997), and compariso of assemblages on these structures with those of rocky shores (Connell and Glasby, 1999; Glasby, 1999). More generally, studies on artificial substrata have contributed to the development and refinement of ecological models to explain patter in natural habitats (e.g., Sutherland, 1974; Sutherland and Karlson, 1977; Anderson, 1998). In contrast, much less is known about patter and processes of intertidal assemblages on artificial substrata. The present study is part of a research programme on the ecology of epibenthic assemblages on hard substrata, including both rocky shores and artificial reefs, in the northwest Mediterranean. Here we focus on the effects of limpets in midshore habitats provided by artificial reefs. Understanding grazing on artificial substrata is important for comparative purposes, since this is one of the most inteively studied and better understood processes on rocky shores. Many studies from different geographical areas have shown that molluscan herbivores can have profound effects on the structure of assemblages in natural habitats (Underwood, 1980; Lubchenco and Gaines, 1981; Hawki and Hartnoll, 1983). In addition to the direct effects documented by these studies, intertidal gastropods can also exert a variety of indirect effects. For example, they can prevent the monopolization of the substratum by ephemeral algae thereby enhancing the establishment of other algae and invertebrates (reviewed in Sousa and Connell, 1992). Patella aspera Roeding and Patella rustica L. are the most common herbivores in midshore and lowshore habitats of rocky coasts in the Mediterranean. These gastropods exhibit different patter of vertical distribution, with P. rustica being more abundant higher on the shore and P. aspera dominating lower down, but with exteive areas of

3 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) overlap at heights on the shore between 0.1 and 0.2 m above the mean-low-water-level (Menconi et al., 1999). Previous experiments (Benedetti-Cecchi and Cinelli, 1993, 1997; Benedetti-Cecchi et al., 1996; Benedetti-Cecchi, 2000) revealed the important role of these grazers in regulating patter of colonization in disturbed patches. The removal of limpets resulted in the monopolization of the substratum by filamentous algae, whereas in the presence of grazers succession proceeded with the establishment of the fleshy red alga Rissoella verruculosa (Bertolini) J. Agardh, barnacles and the Cyanophyta Rivularia spp. More recently, large-scale studies (employing scales similar to those of the present work) have shown coiderable spatial variability in the effects of limpets among shores te to hundreds of kilometres apart (Benedetti-Cecchi et al., in press). These experiments were carried out on rocky shores adjacent to the artificial structures studied in the present paper and the two species of limpets, which were found at deities similar to those here reported (see Section 3), were coidered as a guild and excluded by mea of cages (Benedetti-Cecchi et al., in press). The focus of this paper is on the effects of P. aspera and P. rustica on assemblages of algae and barnacles developing on artificial reefs in the northwest Mediterranean. The primary aim of this study was to test the hypothesis that removing limpets from these manufactures led to changes in the structure of assemblages similar to those observed on rocky shores. Furthermore, we examined whether there were incoistencies in the effects of limpets at spatial scales comparable to those investigated on rocky shores. These hypotheses were tested with a multifactorial experiment involving the orthogonal manipulation of the presence/ absence of the two species of limpets. The experiment was repeated at different locatio (te of kilometres apart), on different reefs within each location (a few kilometres apart), and using replicate boulders within reefs (te to hundreds of metres apart) as the experimental units. This experiment also allowed us to test the null hypotheses that P. rustica and P. aspera had similar effects on assemblages, and that one species had no influence on the distribution of the other on the artificial reefs (these hypotheses have not been tested yet on rocky shores). 2. Materials and methods 2.1. Study site This study was done at two exposed locatio (Carrara, N, E and Livorno, N, E), on the northwest coast of Italy, between June 1997 and February The two locatio were about 70 km apart and were characterised by the presence of industrial developments and marinas. Two artificial reefs, about 4 km apart, were used at each location (referred to as Reef 1 and Reef 2 from north to south at each location, respectively). These reefs were m long and run parallel to the coastline experiencing intee wave action due to western winds (from southwest to northwest). The reefs were made of traplanted carbonatic boulders, with their longer axis ranging from 1 to 3 m. Assemblages on these boulders were qualitatively similar to those occurring in midshore and lowshore habitats of rocky coasts in the area (Benedetti-Cecchi and

4 4 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Cinelli, 1993, 1997; Benedetti-Cecchi et al., 1996; Menconi et al., 1999; Benedetti- Cecchi, 2000). The most common algae were encrusting corallines, the brown crust Ralfsia verrucosa (Areschoug) J. Agardh, Cyanophyta of the genus Rivularia, and erect algae such as Rissoella verruculosa (Bertolini) J. Agardh, Nemalion helmintoides (Velley) Batters, Porphyra leucosticta Thuret, the articulated corallines Corallina elongata Ellis and Solander and Haliptilon virgatum (Zanardini) Garbary and Johaen, the coarsely branched Laurencia obtusa (Hudson) Lamouroux and Chondria spp. (De Notaris) De Toni, and the filamentous Polysiphonia spp. and Ceramium spp. The most abundant sessile invertebrates were the barnacles Chthamalus montagui Southward and Chthmalus stellatus (Poli), while the main herbivores were the limpets Patella aspera, Patella rustica and the snail Osilinus turbinatus (Von Born); these herbivores were distributed at heights on the shore ranging from to 0.4 m with respect to mean-low-water-level Experimental desig and analysis of data Twelve boulders with the longer axis no less than 2 m in length were selected randomly on the seaward side of each reef. These boulders were numbered with marine epoxy for identification and randomly allocated to four treatments with three replicates each. Treatments were: (1) control, where all limpets had been left in place ( 1 Pa 1 Pr), (2) removal of P. aspera (2Pa 1 Pr), (3) removal of P. rustica ( 1 Pa 2 Pr), and (4) removal of both species (2Pa 2 Pr). Limpets were removed by hand with the aid of a screw driver; boulders were searched thoroughly at the beginning of the experiment and every 3 4 weeks thereafter to maintain the experimental conditio. The percentage cover of sessile organisms and the deity of mobile grazers were assessed after 4 and 8 months using quadrats of cm in size. Estimates of percentage cover were obtained visually by subdividing the quadrat in cm sub-quadrats (Dethier et al., 1993; Benedetti-Cecchi et al., 1998). Deities of grazers were expressed as the number of animals present in each quadrat. Sampling was restricted to the seaward side of the boulders, because only this side provided a habitat comparable to that of rocky shores (authors personal observation). Three quadrats were placed randomly on each boulder at each sampling occasion. To avoid re-sampling the same quadrats and to maintain temporal independence in the data, the first set of replicates was marked with epoxy putty so that they were avoided when boulders were sampled the second time Analysis of data Data from the two locatio were analysed separately, since the deities of the two species of limpets and the assemblages were markedly different. A five-factor mixed model analysis of variance was used to test hypotheses about the effects of limpets on artificial reefs. Factors were: P. aspera (fixed and orthogonal), P. rustica (fixed and orthogonal), Time (fixed and orthogonal), Reef (random and orthogonal), and Boulder (random, nested within the interaction Reef 3 P. aspera 3 P. rustica). The respoe variables for these analyses were: (1) encrusting coralline algae,

5 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) (2) Cyanophyta, (3) articulated coralline algae, (4) filamentous algae, (5) Ralfsia verrucosa, (6) Rissoella verruculosa, and (7) Chthamalus spp. The same model of analysis was used to determine the efficacy of the manipulation and to test for the effects of one species of limpet on the deity of the other. In these tests, the factor corresponding to the species of limpet which was also the respoe variable in the analysis, was termed Removal and tested for the efficacy of the experimental manipulation. Cochran s C-test (Winer, 1971; Underwood, 1997) was used to check the assumption of homogeneity of variances. In some cases it was necessary to traform the data (square root or logarithmic scale) to meet this assumption. Pooling procedures were also used when appropriate, according to Winer (1971). Student Newman Keuls tests (SNK) were used for a posteriori compariso of the mea. 3. Results 3.1. Deity and size of limpets Manual removal was effective in reducing the deity of P. aspera in the appropriate treatments at Carrara, although it was impossible to maintain boulders completely free of these herbivores (Fig. 1A,C); the analysis detected a significant main effect of Removal (F , P, 0.05, MS ), indicating that the manipulation was 1,1 Reef3Removal coistent across treatments. At Livorno, no significant effect of the manipulation of P. aspera was disclosed by the analysis, but a seible decrease in its deity, in particular on Reef 2, can be noticed by ipection of the graphs (Fig. 1B,D). The efficacy of removing P. rustica changed significantly from time to time and from reef to reef at Carrara (analysis on ln(x 1 1) traformed data, C , P. 0.05; Reef 3 Removal 3 Time: F , P, 0.05, MS and Fig. 1E,G). 1,16 Boulder3Time SNK tests within this interaction, however, indicated that manual removal of P. rustica significantly reduced the deity of this species compared to unmanipulated boulders on all reefs at both sampling occasio, but on Reef 1 at Time 1. In contrast, the effectiveness of the manipulation of P. rustica at Livorno was coistent in time and between reefs (Fig. 1F,H), resulting in a significant main effect of the Removal (analysis on ln(x 1 1) traformed data, C , P. 0.05; F , 1,16 P, 0.01, MS ). Boulder At both locatio the analysis disclosed a large heterogeneity among boulders in the deity of P. aspera (Carrara: F , P, 0.001, MS ; Livorno: 16,96 Residual F , P, 0.001, MS ) and P. rustica (Carrara: F , 16,96 Residual 16,96 P, 0.001, MS ; Livorno: F , P, 0.001, MS ). Residual 16,96 Residual The size of P. aspera was not affected by the manual removal and was variable among boulders at both locatio (Carrara: F , P, 0.05, MS ; 16,96 Residual Livorno: F , P, 0.001, MS ). Furthermore the analysis indi- 16,96 Residual cated as significant the effect of main term Time at Carrara (F , P, 0.05, 1,1 MS ). Reef3Time Conversely the size of P. rustica was different between removal and control boulders

6 6 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Fig. 1. Mean deity ( 1 S.E., n 5 9) of Patella aspera and Patella rustica in the different treatments before the initiation of the experiment, and after 4 and 8 months. Data are values from three replicate quadrats pooled across three replicate boulders at each time.

7 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) at Carrara, but the effects of the manipulation were not coistent through time and between reefs (Reef 3 Removal 3 Time: F1, , P, 0.05, MSResidual ); SNK tests revealed that larger specime always dwelled on boulders where this species was left at natural deities, except for Reef 1 after 4 months from the initiation of the experiment, where the size did not differ between treatments. At Livorno the analysis detected a significant effect of the interaction Removal 3 P. aspera 3 Time (F1, , P, 0.05, MSTime3Reef3Removal3P. aspera ), suggesting that the size of P. rustica was affected by the manipulation, which effects varied through time and with the removal of the other species. Anyway, the SNK test indicated that larger individuals were on the boulders where this species was left untouched, irrespectively for the presence or absence of P. aspera, at both times of sampling Effects on algae and barnacles Limpets had no effect on the percentage cover of Cyanophyta (Fig. 2A D), which were abundant at Livorno and nearly absent at Carrara. At the former location the abundance of Cyanophyta also varied among boulders and between reefs (Table 1); variability was not coistent through time at the smaller spatial scale. P. aspera had significant effects on the percentage cover of encrusting coralline algae at Livorno, but they were variable between reefs (Fig. 2F,H and Table 1). The removal of P. aspera resulted in a significant reduction in the percentage cover of encrusting coralline algae on Reef 1, while the opposite occurred on Reef 2 (Fig. 2F,H and Table 1). The pattern observed on Reef 1 probably reflected initial differences in the abundance of these algae among boulders assigned to different treatments (Fig. 2F). Finally, the analysis detected a large variability among boulders (Table 1). Encrusting coralline algae were nearly absent at Carrara and their percentage cover varied through time (F , P, 0.05, MS ). 1,1 Reef3Time In contrast to the patter described above, Ralfsia verrucosa was abundant on reefs at Carrara while it was poorly represented at Livorno (Fig. 2I,N). At the former location, both the limpets affected the abundance of this species, but their effects were variable between reefs, resulting in a significant Reef 3 P. rustica 3 P. aspera interaction (Table 1). SNK tests within this interactio did not indicate any effect of limpets (Table 1). Also the interaction P. aspera 3 Time was significant (Table 1) and SNK tests showed that the removal of this species negatively affected the percentage cover of R. verrucosa at Time 1, while it had no effect at Time 2. At Livorno, the abundance of R. verrucosa was higher on boulders where P. aspera was removed (especially on Reef 2), but the analysis did not disclose any significant effect of the manipulation of this species. The percentage cover of R. verrucosa was variable among boulders (F 5 P, 0.01, MS ). 16,96 Residual P. aspera and P. rustica interactively affected the percentage cover of articulated coralline algae at Carrara, but patter changed from reef to reef and from time to time (Fig. 3A,C and Table 2). This was shown by the significant Reef 3 P. rustica 3 P. aspera 3 Time interaction (Table 3). SNK tests within this interaction, however, showed few significant differences and those that were significant were not coistent across

8 8 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Fig. 2. Effect of different combination of absence/ presence of Patella aspera and Patella rustica on mean percentage cover of encrusting algae ( 1 S.E., n 5 9) as a function of time. Data are values from three replicate quadrats pooled across three replicate boulders at each time.

9 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) Table 1 ANOVAs on the effects of the removal of limpets, Reef, Boulder and Time on the percent cover of encrusting algae Source of d.f. Cyanophyta Encrusting corallines Ralfsia verrucosa variation Livorno Livorno Carrara MS F MS F MS F Reef 5 Re ** P. rustica 5 Pr P. aspera 5 Pa Time 5 Ti Re 3 Pr Re 3 Pa * Re 3 Ti Pr 3 Pa Pr 3 Ti Pa 3 Ti * Re 3 Pr 3 Pa * Re 3 Pr 3 Ti Re3 Pa 3 Ti a Pr3 Pa 3 Ti Re3 Pr3 Pa 3 Ti Boulder (Re3 Pr 3Pa) *** Boulder (Re3 Pr 3Pa) 3 Ti *** Residual Cochran s test C , P C , P C , P Traformation ln (x 1 1) ln (x 1 1) None SNK tests Encrusting corallines Ralfsia verrucosa Re 3 Pa Re 3 Pr 3 Pa d.f. 5 1,16; S.E d.f 5 1,16; S.E Reef 1: 1 Pa.2Pa Reef 1 1 Pa: 1 Pa 52Pr 2 Pa: 2 Pr 51Pr Reef 2: 2 Pa.1Pa Reef 1 1 Pr: 1 Pa 52Pa 2 Pr: 2 Pa 51Pa Reef 2 1Pa: 1 Pr5 2Pr 2 Pa: 2Pr 51Pr Reef 2 1 Pr: 2 Pa 51Pa 2 Pr: 2 Pa 52Pa Pa 3 Ti d.f. 5 1,16; S.E Time 1 Time 2 1 Pa.2Pa 2 Pa 51Pa a Re 3 Pa 3 Ti has been eliminated as no significant at P *P, 0.05; **P, 0.01; ***P, , not significant. Analysis relating to locatio where the percentage cover of these algae was very low are not reported in this table, but relevant results are reported in the text (see Section 3). reefs or times. For these reaso, this interaction was not coidered likely to invalidate compariso of other sources of variation. The other significant interaction was that between Reef and P. rustica (Table 2). The removal of P. rustica resulted in a significant increase in cover of articulate coralline algae on Reef 1, but not on Reef 2 (Fig. 3A,C and Table 2); this effect was more evident at Time 2. The data about articulated corallines algae from Livorno were not analysed, since they were in practice absent at this location.

10 10 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Fig. 3. Effect of different combination of absence/ presence of Patella aspera and Patella rustica on mean percentage cover of erect algae ( 1 S.E., n 5 9) as a function of time. Data are values from three replicate quadrats pooled across three replicate boulders at each time.

11 Table 2 ANOVAs on the effects of the removal of limpets, Reef, Boulder and Time on the percent cover of erect algae Source of d.f. Articulated corallines Filamentous algae Rissoella verruculosa variation Carrara Carrara Livorno Livorno MS F MS F MS F MS F Reef 5 Re P. rustica 5 Pr P. aspera 5 Pa ** Time 5 Ti a Re 3 Pr * Re 3 Pa Re 3 Ti b * Pr 3 Pa * Pr 3 Ti Pa 3 Ti Re 3 Pr 3 Pa Re 3 Pr 3 Ti Re3 Pa 3 Ti Pr3 Pa 3 Ti * Re3 Pr3 Pa 3 Ti ** Boulder (Re3 Pr 3Pa) * * ** Boulder (Re3 Pr 3Pa) 3 Ti * Residual Cochran s test C , P C , P C , P C , P Traformation ln (x 11) ln (x 11) None ln (x 1 1) SNK tests Articulated corallines Filamentous algae (Livorno) Rissoella verruculosa Re 3 Pr Pr 3 Pa Pr 3 Pa 3 Ti d.f. 5 1,16; S.E d.f. 5 1,16; S.E d.f. 5 1,16; S.E Reef 1: 2 Pr.1Pr 1 Pa: 2 Pr 51Pr Time 1 Time 2 Reef 2: 1 Pr 52Pr 2 Pa: 2 Pr 51Pr 1 Pa: 2 Pr 51Pr 1 Pa: 2 Pr 51Pr 1Pr: 2Pa 51Pa 2Pa: 2 Pr5 1Pr 2 Pa: 1 Pr.2Pr 2Pr: 2Pa.1Pa 1 Pr: 2 Pa 51Pa 1Pr: 2Pa.1Pa 2 Pr: 2 Pa 51pa 2 Pr: 1 Pa.2Pa a See Section 3 for details. b Re 3 Ti has been eliminated from the analysis, as no significant at P *P, 0.05; **P, 0.01; ***P, , not significant. Analysis relating to locatio where the percentage cover of these categories of algae was very low are not displayed in this table, but relevant results are reported in the text. F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000)

12 12 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Table 3 ANOVAs on the effects of the removal of limpets, Reef, Boulder and Time on the percent cover of Chthamalus spp. Source of d.f. Carrara Livorno variation MS F MS F Reef 5 Re P. rustica 5 Pr P. aspera 5 Pa Time 5 Ti Re 3 Pr *** Re 3 Pa Re 3 Ti * Pr 3 Pa * Pr 3 Ti Pa 3 Ti Re 3 Pr 3 Pa 1 a Re 3 Pr 3 Ti 1 b ** Re 3 Pa 3 Ti Pr 3 Pa 3 Ti Re 3 Pr 3 Pa 3 Ti * Boulder (Re 3 Pr 3 Pa) *** * Boulder (Re 3 Pr 3 Pa) 3 Ti Residual Cochran s test C , P C , P Traformation ln (x 1 1) ln (x 1 1) SNK tests Carrara Re 3 Pr 3 Ti d.f. 5 1,16; S.E Time 1 Time 2 Reef 1: 2 Pr.1Pr Reef 1: 1 Pr 52Pr Reef 2: 2 Pr 51Pr Reef 2: 2 Pr.1Pr Livorno Pr 3 Pa Re 3 Pr d.f. 5 1,16; S.E d.f. 5 1,16; S.E Pa: 2 Pr.1Pr 2 Pa: 1 Pr 52Pr Reef 1: 2 Pr.1Pr 1 Pr: 2 Pa.1Pa 2 Pr: 1 Pa 52Pa Reef 2: 1 Pr.2Pr a Re 3 Pr 3 Pa has been eliminated from the analysis as no significant at P b See text for details. *P, 0.05; **P, 0.01; ***P, , not significant. P. aspera significantly reduced the percentage cover of filamentous algae at Carrara (Fig. 3E,G and Table 2). This effect was not evident at Time 2 because there were few filamentous algae in all plots on that sampling occasion (Fig. 3E,G); anyway, the analysis indicated as significant the effect of the main terms P. aspera and Time, but not that of their interaction (Table 2). There were significant differences among boulders in the percentage cover of filamentous algae that changed from time to time (Table 2). At the other location, both the species of limpets affected the cover of filamentous algae (Fig. 3F,H), resulting in a significant P. rustica 3 P. aspera interaction. SNK test

13 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) could not identify a clear pattern for this interaction, but suggested that in absence of P. rustica the percentage cover of these algae was higher when also was P. aspera removed. Furthermore, there were large differences in the abundance of filamentous algae from reef to reef and from boulder to boulder, which were not coistent through time in the former case. Rissoella verruculosa was absent at Carrara, while it occurred on the reefs at Livorno (Fig. 3I N). The effects of the two species of limpets at Livorno were complex and interactive, also changing from time to time and resulting in a significant P. rustica 3 P. aspera 3 Time interaction (Table 2). No effects of limpets were revealed at Time 1, while at Time 2 the removal of one species of limpets resulted in a significant increase in the percentage cover of Rissoella, but only if the other species was present. In contrast, removing both species produced a decline in cover of the alga (Fig. 3L,N and Table 2). The percentage cover of Rissoella changed significantly among boulders (Table 2). Barnacles were more abundant at Livorno than at Carrara (Fig. 4). At the latter location, P. rustica and P. aspera affected the percentage cover of Chthamalus spp. but these effects were not coistent across reefs and times of sampling, according to the significant Reef 3 P. rustica 3 P. aspera 3 Time interaction (Fig. 4A,C and Table 3). As just a few significant differences, not coistent across reefs and times, were pointed out Fig. 4. Effect of different combination of absence/ presence of Patella aspera and Patella rustica on mean deity of Chthamalus spp. ( 1 S.E., n 5 9) as a function of time. Data are values from three replicate quadrats pooled across three replicate boulders at each time.

14 14 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 by SNK tests, lower order significant interactio were further coidered. This was the case of the interaction Reef 3 P. rustica 3 Time, which indicated that the removal of P. rustica increased the percentage cover of barnacles on Reef 1 at Time 1, and on Reef 2 at Time 2 (Fig. 4 and Table 3). At Livorno also, the analysis suggested an interactive effect of both the species of limpets on the percentage cover of Chthamalus spp. (Fig. 4B,D and Table 3). The removal of P. rustica and P. aspera had a positive effect on the abundance of these barnacles, but only when the other species was left untouched; there were no further effects of removal of a species in absence of the other (SNK test, Table 3). Furthermore, the interaction Reef 3 P. rustica was significant and the SNK test indicated that the removal of this limpet increased the abundance of barnacles on Reef 1, while the opposite occurred on Reef 2. The percentage cover of barnacles varied among boulders at both locatio (Table 3). 4. Discussion The results of this study indicate that limpets can have strong and relatively coistent effects on filamentous algae on artificial reefs, whereas interactio with other members of assemblages were complex and variable. This complexity was revealed by incoistencies in the effects of the limpets at different spatial scales and through time. Our data cannot test for the effects of grazing at the largest spatial scale, between locatio, since they are totally confounded with differences between assemblages: Cyanophyta, encrusting coralline algae and Rissoella were abundant at Livorno but nearly absent at Carrara, whereas the opposite was true for articulate coralline algae and Ralfsia; only filamentous algae showed a similar abundance at the two study locatio. Also the deity of limpets was different at the scale of te of kilometers, being the two species more abundant at Carrara. Within locatio, effects due to one or the other species of limpets, additive or multiplicative effects of these grazers and interactio between limpets and time, were common. Previous studies in the northwest Mediterranean have shown that limpets can have strong effects on filamentous algae both in mid-shore and low-shore habitats on rocky coasts (Benedetti-Cecchi and Cinelli, 1993; Benedetti-Cecchi et al., 1996; Benedetti- Cecchi, 2000). Similar effects have been documented in the present study on artificial reefs, where filamentous algae rapidly colonised boulders maintained at reduced deities of grazers, and in particular when P. aspera was removed. Recent experiments on rocky shores, conducted in parallel to the present one, have shown incoistencies in the effects of limpets on filamentous algae among locatio te to hundreds of kilometres apart, at scales comparable to those investigated in the present study (Benedetti-Cecchi et al., in press). On rocky shores, however, incoistencies in space of the effects of limpets reflected true variation in their foraging activity, while on artificial reefs they were likely the product of vagaries in the recruitment of algae. In addition, while studies on rocky shores have documented weak effects of limpets at Punta Bianca (a location very close to Carrara) and strong effects at Calafuria

15 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) (a location very close to Livorno) (Benedetti-Cecchi and Cinelli, 1993; Benedetti- Cecchi, 2000; Benedetti-Cecchi et al., in press), the largest effects of limpets on artificial reefs occurred at Carrara. The opposite direction of these patter strongly suggest that the processes driving large-scale variation in the effects of limpets on rocky shores are different from those operating on artificial reefs. Differences in the methodology used to manipulate limpets (fences were used to exclude limpets on rocky shores while manual removal was used in the present study), might also account for some of these incoistencies. Collectively, the results above suggest that filamentous algae are largely affected by limpets both on rocky shores and artificial reefs. Grazing on artificial reefs, however, appeared more coistent than grazing on rocky shores, provided that the filamentous algae were able to colonize. Further experiments are needed to determine whether or not these patter reflect real differences between natural and artificial habitats in the processes accounting for large-scale variation in grazing. Our results show that of the two species of limpets manipulated, P. aspera had major effects on filamentous algae. Although less abundant than P. aspera, P. rustica was 22 common on control boulders at deities in the range of ind. 100 cm. Differences between removal and control boulders were maintained throughout the experiment by manual removals, despite spatial and temporal variability in the efficacy of the manipulation. The proportional reduction of P. rustica by manual removal was, however, larger than that achieved with P. aspera (Fig. 1). These patter suggest that the lack of any effect of P. rustica at Carrara on filamentous algae is not an artifact of the experimental procedure. A tentative explanation for the observed patter is that P. rustica did not graze for long enough in the experimental plots to keep fast-growing organisms, such as the filamentous algae, under control (although this species did affect other members of the assemblage, as discussed below). Possibly, individuals of P. rustica tended to move up-shore during their foraging excursio to graze on different assemblages of microalgae. Unfortunately, studies on rocky shores do not assist in the interpretation of these results, since no attempt has been done to separate the effects of the two species of limpets in the natural habitat. Indirect effects of limpets on rocky shores have been documented in several studies (Lubchenco and Menge, 1978; Sousa, 1979; Underwood et al., 1983; Van Tamelen, 1987; Kim, 1997). These effects often involve a chain of interactio with limpets preventing the monopolization of the substratum by fast-growing algae, thereby allowing the colonization of other organisms that require primary substratum to become established. By removing filamentous algae, limpets indirectly facilitated the establishment of Rissoella and, apparently, barnacles, on rocky shores at Livorno (Benedetti- Cecchi and Cinelli, 1993; Benedetti-Cecchi, 2000). P. aspera and P. rustica interactively affected the percentage cover of Rissoella on artificial reefs at Livorno. At Time 2 the removal of one or the other species resulted in an increase in abundance of Rissoella compared to unmanipulated boulders. Therefore, there was no evidence for the occurrence of positive indirect effects of limpets on Rissoella when the two species of grazers were present simultaneously. Possibly, the lack of indirect effects was a coequence of the low cover of filamentous algae on both reefs at Livorno. In the absence of the inhibitory effect of filamentous algae (see Benedetti-Cecchi, 2000), there

16 16 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 was no way for Rissoella to benefit from the presence of limpets and negative effects predominated. In contrast, when both species of limpets were removed and colonization by filamentous algae was, to some extent, successful (at Time 2), the cover of Rissoella declined compared to boulders where only one species of limpets was present (Fig. 3H,N). These outcomes suggest that under moderate grazing, and in the presence of filamentous algae, limpets can generate patter like those expected in the presence of indirect effects. Differences among treatments were, however, negligible and probably biologically irrelevant. Limpets are known to affect the abundance of barnacles on rocky shores in different ways. For example, they can enhance the cover of barnacles indirectly, by preventing the monopolization of the substratum by algae (see above). These patter have been described for rocky shores at Calafuria (Benedetti-Cecchi, 2000) and in other regio around the world (Hawki, 1983; Underwood et al., 1983; Dungan, 1986; Van Tamelen, 1987). Other studies, however, have shown that limpets can reduce the cover of barnacles by grazing newly settled cyprids or by bulldozing juveniles from the substratum (Dayton, 1971; Branch, 1975; Denley and Underwood, 1979). Barnacles were mostly affected by P. rustica in the present study and effects were largely negative. On Reef 2 at Livorno, however, the removal of P. rustica resulted in a decline in cover of barnacles indicating a positive effect. As already discussed for Rissoella (see above), this might reflect an indirect effect mediated by filamentous algae. This interpretation is, however, only tentative, because filamentous algae never monopolized the substratum on Reef 2. At present, we lack a better explanation for the observed patter. Studies on rocky shores have shown that grazing by limpets may be important for the persistence of encrusting algae that otherwise would be replaced by erect species (Paine, 1980; Steneck, 1982). Interactio between limpets and encrusting algae were not so clear on artificial reefs. Removal of P. aspera resulted in an increase in percentage cover of encrusting coralline algae on Reef 2 at Livorno, revealing negative rather than positive effects of grazing. Anyway, the effect of P. aspera on these algae must be interpreted with caution, since the differences between presence/ absence of this limpet were present from the beginning of the experiment. The opposite was observed on Reef 1 at Livorno, but this probably reflected the fact that, by chance, boulders with exteive cover of encrusting corallines were assigned to the control treatment at the beginning of the experiment, while boulders supporting few encrusting coralline algae were assigned to the removal treatment (see Fig. 2F). Thus, there was no evidence of positive effects of grazing on encrusting corallines from these data. There were both positive and negative effects of limpets on Ralfsia at Carrara. For example removal of P. aspera resulted in a decline in cover of Ralfsia at Time 1. This probably reflected the increase in cover of filamentous algae on boulders where P. aspera was removed, suggesting that indirect effects of grazers through filamentous algae were important for the persistence of Ralfsia on artificial reefs. This pattern was reversed at Time 2, when removal of P. aspera resulted in an increase in cover of Ralfsia, although it was not significant. At this time, however, filamentous algae were nearly absent on experimental boulders so that positive indirect effects of limpets were unlikely. From these results it seems that, on artificial reefs, the net effect of limpets on

17 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) Ralfsia can shift from a positive indirect interaction in the presence of filamentous algae, to a direct negative effect that occurs only in the absence of filamentous algae. The effects of limpets on Ralfsia could not be interpreted in light of the significant Reef 3 P. aspera 3 P. rustica interaction, which reflected the average effects of grazing over the two sampling occasio, since no clear pattern could be identified. Although the magnitude and generality of these interactio is far from clear from our results, they do indicate that on artificial reefs the positive effect of limpets on encrusting algae cannot be simply typified as a chain of interactio involving grazers, encrusting algae and their competitors. Rather, patter are likely to involve complex and yet unexplained interactio among coumers and between coumers and a range of potential resources. In conclusion, this study has shown that grazing by limpets is an important process on artificial reefs, and a wide range of effects of P. aspera and P. rustica have been documented. The collective effects of the two species are, in large part, comparable to those described for rocky shores in the same area. The processes accounting for large-scale variation in grazing seem, however, different between the natural and the artificial habitat. Understanding this variability is important to assess whether artificial reefs can reproduce the same scales of variation, both in terms of patter and processes, observed on rocky shores. Future studies should be directed to understand whether P. aspera and P. rustica display the same range of interactio on rocky shores as those described here (from independent to additive and multiplicative effects). Filling this gap would provide an additional basis to assess the extent to which artificial reefs can be coidered substitutive of natural habitats in terms of relevant ecological processes. Acknowledgements We sincerely thank A. Del Nista and P. Nugnes for their help on field, R.A. Coleman, L. Airoldi and G. Ceccherelli and two anonymous referees for their criticism on the early draft of the manuscript. This work was partially supported by the EC under MAST programme contract MAS3-CT (EUROROCK). [AU] References Ambrose, R.F., Mitigating the effects of a coastal power plant on a kelp forest community: rationale and requirements for an artificial reef. Bull. Mar. Sci. 55, Anderson, M.J., Effects of patch size on colonisation in estuaries: revisiting the species-area relatiohip. Oecologia 118, Anderson, M.J., Underwood, A.J., Effects of gastropod grazers on recruitment and succession of an estuarine assemblage: a multivariate and univariate approach. Oecologia 109, Benedetti-Cecchi, L., Predicting direct and indirect interactio during succession in a midlittoral rocky shore assemblage. Ecol. Monogr. 70 (1), Benedetti-Cecchi, L., Cinelli, F., Early patter of algal succession in a midlittoral community of the Mediterranean sea: a multifactorial experiment. J. Exp. Mar. Biol. Ecol. 169,

18 18 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) 1 19 Benedetti-Cecchi, L., Cinelli, F., Confounding in field experiments: direct and indirect effects of artifacts due to manipulation of limpets and macroalgae. J. Exp. Mar. Biol. Ecol. 209, Benedetti-Cecchi, L., Nuti, S., Cinelli, F., Analysis of spatial and temporal variability in interactio among algae, limpets and mussels in low-shore habitats on the west coast of Italy. Mar. Ecol. Prog. Ser. 144, Benedetti-Cecchi, L., Bulleri, F., Cinelli, F., Deity dependent foraging of sea urchi in shallow subtidal reefs on the west coast of Italy (western Mediterranean). Mar. Ecol. Prog. Ser. 163, Benedetti-Cecchi, L., Bulleri, F., Acunto, S., Cinelli, F., in press. Scales of variation in the effects of limpets on rocky shores in the northwest Mediterranean. Mar. Ecol. Prog. Ser. Branch, G.M., Intraspecific competition in Patella cochlear Born. J. Anim. Ecol. 44, Breitburg, D.L., Development of a benthic subtidal epibenthic community: factors affecting species composition and the mechanisms of succession. Oecologia 65, Carr, M.H., Hixon, M.A., Artificial reefs: the importance of compariso with natural reefs. Fisheries 22, Connell, S.D., Glasby, T.M., Do urban structures influence local abundance and diversity of subtidal epibiota? A case study from Sydney Harbour, Australia. Mar. Environ. Res. 47, Dayton, P.K., Competition, disturbance, and community organization: the provision and subsequent utilization of space in a rocky intertidal community. Ecol. Monogr. 41, Denley, E.J., Underwood, A.J., Experiments on factors influencing settlement, survival and growth of two species of barnacles in New South Wales. J. Exp. Mar. Biol. Ecol. 36, Dethier, M.N., Graham, E.S., Coen, S., Tear, L.M., Visual versus random-point percent cover estimatio: objective is not always better. Mar. Ecol. Prog. Ser. 110, Dungan, M.L., Three-way interactio: barnacles, limpets, and algae in a sonoran desert rocky intertidal zone. Am. Nat. 127, Glasby, T.M., Differences between subtidal epibiota on pier pilings and rocky reefs at marinas in Sydney, Australia. Estuar. Coast. Shelf Sci. 48, Grosberg, R.K., Intertidal zonation of barnacles: the influence of planktonic zonation of larvae on vertical distribution of adults. Ecology 63, Hawki, S.J., Interaction of Patella and macroalgae with settling Semibalanus balanoides (L.). J. Exp. Mar. Biol. Ecol. 71, Hawki, S.J., Hartnoll, R.G., Grazing of intertidal algae by marine invertebrates. Oceanogr. Mar. Biol. Annu. Rev. 21, Keough, M.J., Effects of patch size on the abundance of sessile marine invertebrates. Ecology 65, Keough, M.J., Butler, A.J., Temporal changes in species number in an assemblage of sessile marine invertebrates. J. Biogeogr. 10, Kim, J.H., The role of herbivory, and direct and indirect interactio, in algal succession. J. Exp. Mar. Biol. Ecol. 217, Lubchenco, J., Menge, B.A., Community development and persistence in a low rocky intertidal zone. Ecol. Monogr. 48, Lubchenco, J., Gaines, S.D., A unified approach to marine plant-herbivore interactio. I. Populatio and communities. Annu. Rev. Ecol. Syst. 12, Menconi, M., Benedetti-Cecchi, L., Cinelli, F., Spatial and temporal variability in the distribution of algae and invertebrates on rocky shores in the northwest Mediterranean. J. Exp. Mar. Biol. Ecol. 233, Paine, R.T., Food webs: linkage, interaction strength and community infrastructure. The Third Taley lecture. J. Anim. Ecol. 49, Seaman, W., Buckley, R.M., Polovina, J.J., Advances in knowledge and priorities for research, technology and management related to artificial aquatic habitats. Bull. Mar. Sci. 44, Sousa, W.P., Experimental investigatio of disturbance and ecological succession in a rocky intertidal community. Ecol. Monogr. 49, Sousa, W.P., Connell, J.H., Grazing and succession in marine algae. In: John, D.M., Hawki, S.J., Price, J.H. (Eds.), Plant Animal-Interactio in the Marine Benthos. Clarendon, Oxford, UK, pp Steneck, R.S., A limpet-coralline alga association: adaptatio and defees between a selective herbivore and its prey. Ecology 63,

19 F. Bulleri et al. / J. Exp. Mar. Biol. Ecol. 255 (2000) Sutherland, J.P., Multiple stable points in natural communities. Am. Nat. 108, Sutherland, J.P., Karlson, R.H., Development and stability of the fouling community at Beaufort, North Carolina. Ecol. Monogr. 47, Underwood, A.J., The effects of grazing by gastropods and physical factors on the upper limits of distribution on intertidal macroalgae. Oecologia 46, Underwood, A.J., In: Experiments in Ecology: Their Logical Design and Interpretation Using Analysis of Variance. Cambridge University Press, Cambridge. Underwood, A.J., Denley, E.J., Moran, M.J., Experimental analysis of the structure and dynamics of mid-shore rocky intertidal communities in New South Wales. Oecologia 48, Van Tamelen, P.G., Early successional mechanisms in the rocky intertidal: the role of direct and indirect interactio. J. Exp. Mar. Biol. Ecol. 112, Winer, B.J., In: Statistical Principles in Experimental Design, 2nd Edition. McGraw-Hill, Kogakusha, Tokyo.