Forgot your password?
Document Actions
Arrontes et al 04
MARINE ECOLOGY PROGRESS SERIES
Vol. 277: 117–133, 2004 Published August 16
Mar Ecol Prog Ser
Effect of grazing by limpets on mid-shore species
assemblages in northern Spain
J. Arrontes1,*, F. Arenas2, C. Fernández1, J. M. Rico1, J. Oliveros1, B. Martínez3,
R. M. Viejo3, D. Alvarez1
1
Departamento de Biología de Organismos y Sistemas, Universidad de Oviedo, 33071 Oviedo, Spain
2
Marine Biological Association of the United Kingdom, Prospect Place, Plymouth PL1 2PB, UK
3
Área de Biodiversidad y Conservación, Escuela Superior de Ciencias Experimentales y Tecnología,
Universidad Rey Juan Carlos, 28933 Móstoles, Madrid, Spain
ABSTRACT: Limpets were excluded from barnacle-dominated areas in 2 semi-exposed localities on
the shore of northern Spain. We tested the hypotheses that a macroalgal canopy would develop and
barnacle cover would decrease in exclusion quadrats. Spatial replication considered localities and
sites within localities, while temporal replication considered 2 seasons of removal and 2 start dates
within each locality and season. The experimental methods included exclusion quadrats (using
fences), procedural controls (half fences) and untouched controls. Results supported both hypotheses.
Limpet removal caused exclusion quadrats to change over time more than controls. An algal canopy
dominated by Fucus vesiculosus developed and more barnacles were lost in exclusion quadrats than
in controls. In addition, more trochids (largely Gibbula spp.) and whelks (Thais lapillus) were found
in exclusion quadrats. Site effects were significant for virtually all variables analysed but no effect of
locality, season or date was found. An indirect effect mediated by the development of an algal canopy
appeared to be responsible for the loss of barnacles in exclusion quadrats. This indirect effect might
have been reinforced by an additional indirect effect of algae mediated by increased densities of
whelks under the algal canopy. Altogether, our results agree with those in semi-exposed shores in
other parts of the world but not with results obtained in the west coast of Sweden, which has
contrasting environmental conditions and a different dominant grazer species.
KEY WORDS: Grazing · Community structure · Indirect effects · Patella spp. · Fucus spp. · Spain
Resale or republication not permitted without written consent of the publisher
INTRODUCTION of the grazers, the relative magnitudes of the feeding
rate and the growth rate of the algae (Underwood &
It is well documented that the composition and dy- Jernakoff 1981). At lower tidal levels or in areas with
namics of species assemblages in mid- and high tidal increased shading or wetness conditions, the relative
levels in rocky shores are strongly influenced by the importance of grazing as a structuring force may de-
activity of grazers (Hawkins & Hartnoll 1983). The crease as the growth rates of algae increase. Seasonal
most evident effect of grazing at the high shore is the and spatial variation in the intensity of recruitment is
removal of macroscopic algal species leaving the pri- an additional source of variability of the outcome of the
mary substratum occupied by sessile animals; how- interaction. In experimental shore ecology, there is
ever, the assessment of the effect of grazers is far from ample evidence that timing of manipulations can sig-
being that simple. Even if we are limited to the direct nificantly influence the rate of recovery and the spe-
removal of algae, there is an interaction between graz- cies sequence. Examples abound and cover a great
ing and the effect of physical factors (Underwood variety of habitats such as limpet dominated areas
1980). The final effect depends on the mode of feeding (Hawkins 1981), subtidal kelp forests (Kennelly 1987,
*Email: arrontes@uniovi.es © Inter-Research 2004 · www.int-res.com
118 Mar Ecol Prog Ser 277: 117–133, 2004
Dayton et al. 1992), assemblages of geniculate coral- replication at different time scales. Recent examples of
line algae (Benedetti-Cecchi & Cinelli 1994) or fucoid studies with adequate spatial replication can be found
dominated beds (Kim & DeWreede 1996). The effect of in Benedetti-Cecchi et al. (2000) and Boaventura et
grazers is not, however, limited to the removal of algae al. (2002), the former also with proper temporal
and very often grazers can affect other animal species replication.
through competitive interactions (Underwood 1978, It is obvious that not all species in the grazer assem-
Dethier & Duggins 1984). In this case, care must be blage have the same effect on algae (Underwood &
taken when generalising because the intensity of com- Jernakoff 1981, Underwood 1984). Often, this is the
petition may vary spatially and temporally in response consequence of a single species or a reduced group of
to the densities and identities of competitors and avail- species having a disproportionate effect. Limpets are
ability of food (Underwood 1984). Grazers may also very often the key grazers in mid- and high tidal
affect the rate of succession (e.g. Farrel 1991) or cause levels (Hawkins et al. 1992). In different parts of the
different assemblages to develop (e.g. Anderson & world, other gastropods (Cervin & Åberg 1997, Viejo
Underwood 1997). et al. 1999), chitons (Dethier &Duggins 1984) or even
Apart from direct effects, indirect effects of grazers insect larvae (Robles & Cubit 1981) may play relevant
are common and have been profusely described in roles.
shore environments (e.g. Menge 1995). These indi- The aim of this work was to assess the role of domi-
rect effects can usually be grouped into several nant grazers (limpets) on the structure of species as-
types, one of which is habitat facilitation: the activity semblages in mid-tidal levels of northern Spain at dif-
of one species promoting habitat transformation, ferent spatial and temporal scales. Barnacles are the
which makes it suitable for colonisation by other main space occupiers, in some zones covering up to
species. In some cases, there is a dual effect of the more than 90% of available rock surface. Virtually no
key grazer on a third species, a direct deleterious macroalgae are present and the grazer assemblage is
effect through exploitative or interference competi- limited to a small number of species (Anadón 1983).
tion and an indirect positive effect mediated by habi- Structurally, the system is similar to those found in
tat facilitation (Menge 1995). other parts of the world at the same tidal level
The structuring processes may vary between distant (Stephenson & Stephenson 1972, Raffaelli & Hawkins
areas and, in fact, some studies report an important 1996). The experimental design was intended to make
variability among different geographic zones even broad comparisons of the role of grazers in high and
when they have equivalent species composition (Dethier mid-tidal levels on European shores. The experimental
& Duggins 1988, Boaventura et al. 2002). Obviously, design considered temporal and spatial variability at
broad-scale comparisons of the effect of grazers are 2 different scales: among and within seasons, and
unavoidable in seeking generality of structuring forces among and within localities. Within the framework of
in shore communities. However, making generalisa- an international project (Chelazzi et al. 1998a), identi-
tions from particular studies performed in different cal studies were performed on other European shores
parts of the world at different times may be erroneous with the same experimental design at the same dates:
due to spatial and temporal variability (Foster 1990, Sweden (Lindegarth et al. 2001), Isle of Man, southern
Underwood & Petraitis 1993). Comparisons are often England and southern Portugal. Similar species assem-
difficult due to variable experimental designs and dif- blages but contrasting environmental conditions were
ferent degrees of spatial replication. Even in the case identified in different locations, such as the Isle of
of equivalent designs, variability in time and space Man, southern England and northern Spain. In other
may preclude any comparison (including qualitative cases, both environmental conditions and species dif-
comparisons) of studies performed at single distant fered (e.g. Sweden and northern Spain).
localities at different times. Different localities within
the same shore, although only a few kilometres apart,
may differ, and usually they do, in average orientation, MATERIALS AND METHODS
slope, wave exposure and intensity of recruitment.
Consequently, both density and composition of the Localities and species. The study was carried out
grazer assemblage as well as abundance and growth from May 1996 to October 1998 in 2 localities on the
rates of algae may differ among close locations; hence north shore of Spain, Campiello (43° 33’ N, 6° 24’ W)
the relative importance of grazing as a structuring and Artedo (43° 34’ N, 6° 12’ W). Both localities are
agent. The practical implication is that to cope with moderately exposed to wave action and have a gently
this expected variability, any large scale comparison sloping rock platform with many boulders. The rock
among shores must include replication at different platform in Campiello faces north and is slightly more
spatial scales, from metres to kilometres, and temporal exposed than in Artedo, which faces east. The nature
Arrontes et al.: Effects of grazing by limpets 119
of the rock is the same in both localities, largely replicated at different sites. At the lower part of the
quartzite with some veins of slate. The maximum tidal barnacle-dominated zone, 8 experimental sites inter-
range of spring tides is around 4.3 m. Intertidal com- spersed along the intertidal platforms in Artedo and
munities are very similar in the 2 localities and have 8 in Campiello were selected in late May 1996, before
been described elsewhere (Anadón & Niell 1981, the start of the experimental manipulations. Sites were
Anadón 1983, Arrontes 1993). Briefly, the high shore around 15 m long, at least 30 m apart and had enough
between roughly 3 and 1.5 m above Lowest Astronom- positions to install nine 50 × 50 cm experimental
ical Tide (LAT) is dominated by barnacles and grazing quadrats (see below). Criteria for the selection of the
gastropods. Primary space is largely occupied by 2 quadrats were (1) an inclination between 0 and 45°,
species of barnacles, Chthamalus montagui and C. (2) simple topography and absence of large cracks,
stellatus, small patches of the lichen Lichyna pygmaea crevices or pools, (3) cover of barnacles (Chthamalus
and ephemeral blue green algae. Two species of spp.) above 40%, (4) being at least 1 m apart from the
limpets, Patella depressa and P. vulgata, coexist in this nearest quadrat and (5) absence of perennial macroal-
zone. The relative abundance of both species varies in gae and less than 5% cover of ephemerals. No distinc-
the tidal range and along the shore. Other grazers are tion between stable bedrock and large boulders was
trochids (Gibbula spp. and Monodonta lineata) and made. Temporal variability also considered 2 scales,
small littorinids (species of the Littorina saxatilis group start season (summer and winter) and 2 start dates
and Melaraphe neritoides). In both localities, the within each season, which were different for each
predatory whelk Thais (= Nucella) lapillus is present at locality. Summer start dates were each of the 4 con-
low densities. Below the grazer-dominated zone, secutive spring tides in June–July 1996. Two start
between 1.5 and 0.75 m above LAT, fucoids monopo- dates were randomly assigned to each locality. Winter
lise the space (Fucus vesiculosus and F. serratus in start dates were the 4 spring tides in December
Campiello but only F. vesiculosus in Artedo). The 1996–January 1997.
lower tidal levels are dominated by Bifurcaria bifur- At each locality and date, 2 sites of the 8 previously
cata, Chondrus crispus and Himanthalia elongata (the selected were randomly chosen and 3 replicates of
latter only in Artedo). each grazing treatment were randomly allocated to
Experimental design. Three grazing treatments each of the 9 quadrats previously selected at each
were considered. Limpets were removed and excluded site. The experiment involved a total of 144 quadrats
from 50 × 50 cm quadrats, hereafter named exclusion (2 localities × 2 seasons × 2 start dates × 2 sites × 3 graz-
quadrats (E), using fences 5 cm high made of plastic ing treatments × 3 replicates). With this design, exper-
coated iron mesh (with a mesh size of 1 cm) attached to imental factors include grazing treatment (G, fixed)
the rock with stainless steel screws and plastic plugs. with 3 levels (E, PC and C), locality (L, random) with
Where the fences did not exactly fit the substratum, 2 levels (Campiello and Artedo), season (Se, fixed)
strips of artificial grass were introduced between the with 2 levels (summer and winter), start date (D, ran-
rock and the mesh to prevent small limpets passing dom), nested in the interaction L × Se and with 2 start
under the fences. Artefacts might occur because of dates per locality and season, and site (Si, random),
alterations of water flow within the exclusion quadrats nested in date and with 2 sites per date at each locality
or increased shading and wetness of the rock surface. and season.
Procedural controls (PC) to test for potential artefacts Two hypotheses were derived from the general mo-
due to barriers consisted of partial fences. Quadrats del that limpets influence the structure of assemblages:
were fenced at the corners, leaving an open space of (1) algal cover increases in exclusion quadrats when
25 cm in the middle of each side. At several randomly compared with both types of controls (C = PC < E) and
chosen corners, strips of artificial grass were in- (2) barnacle cover decreases in exclusion quadrats
corporated. Limpets could move freely in and out of when compared with controls (C = PC > E). The unde-
the quadrats. The environmental conditions in the sirable effect of fences is analysed by testing the
quadrats were more similar to those within the exclu- hypothesis that controls differ (C ≠ PC), both for algae
sion quadrats than to those on natural surfaces (but see and barnacle cover. Temporal variability in the effect
comments and criticism on partial barriers as proce- of limpet removal may be detected after significant
dural controls by Johnson [1992] and Benedetti-Cecchi tests for the interactions of grazing with season (G ×
& Cinelli [1997]). Finally, control quadrats (C) were Se) and grazing with date [G × D(Se × L)]. The spatial
marked with 2 screws on opposite corners and left heterogeneity may be detected after significant tests
otherwise untouched. for the interactions of grazing with locality (G × L) and
To cope with spatial heterogeneity, the experiment grazing with site {G × Si[D(Se × L)]}. Because the re-
was replicated in 2 localities (Campiello and Artedo). maining terms in the model were not directly related to
Within each locality, the grazing treatments were the hypotheses, they were ignored in the ‘Discussion’.
120 Mar Ecol Prog Ser 277: 117–133, 2004
Height above LAT, orientation and slope were re- in cover in individual quadrats [100 × (initial cover –
corded for each quadrat. In addition, an estimation of final cover)/initial cover]. Apart from analyses on the
roughness of the quadrats was obtained by loosely lay- effect of limpet removal on the global assemblage of
ing a thin chain with 4 mm links on the 2 diagonals. algae and barnacles, additional analyses tested the
The excess in length of the sum of the 2 diagonals esti- success of fences in excluding limpets and variations in
mated with the chain in relation to the sum of the true the abundance of trochids.
diagonals of a 50 × 50 cm quadrat was considered an With this design, tests for grazing as a main effect
estimate of the roughness of the rock. and some tests for interaction had a reduced power
Sampling. The experiment ran from June 1996 to (df = 2, 2). Different experimental designs might pro-
January 1998. Proportions of Patella vulgata and P. vide more powerful tests for these factors, but the work
depressa were estimated at the start of manipulations reported here was part of a larger experiment and
from animals removed from a variable number of therefore constrained by the need for a common de-
exclusion quadrats in each locality. Quadrats were sign (Chelazzi et al. 1998a). We pooled non-significant
sampled during low spring tides with the point interactions (p > 0.25) involving random factors to
method using a grid made of double thread to avoid increase the power of the tests involving the grazing
parallax errors and with 49 regularly spaced points. treatment. The sequential decision pooling procedure
Samples were taken at the time of the manipulations, described in Winer (1971) was used. Preliminary tests
15 d later, once a month during the first 3 months and were performed to check if the interaction Grazing ×
then every 3 months until January 1998. To follow Date could be eliminated from the model and pooled
up colonisation by Fucus spp., its cover in exclusion with Grazing × Site. Then, we tested the significance of
quadrats was sampled on 2 additional dates, March Grazing × Locality × Season and Grazing × Locality
and October 1998. Primary and secondary cover of all against the pooled error term. If not significant, the
organisms present in the quadrats was recorded and tests for Grazing × Season and Grazing against the
transformed to percentages. Present but not recorded new pooled error term would have considerably
organisms were assigned a cover of 1%. When pos- increased the power (df = 2, 28). Cochran’s test was
sible, organisms were identified to species in the field, used to test for heterogeneity of variances. When
though some taxonomically difficult groups were required, data were transformed to meet the assump-
always assigned to higher categories (e.g. Cerami- tion of homogeneity of variance.
ales). The number of trochids and limpets, with no The relationship between the proportion of Fucus
distinction of species, was also recorded. Limpets cover and height on the shore as well as the effects of
entering the exclusion quadrats were counted and locality (random factor) and season (fixed) were fitted
removed. by a logistic regression using the SAS Macro program
Data analysis. Hypotheses were tested using analy- GLIMMIX, which iteratively uses the SAS MIXED
sis of variance (ANOVA). The existence of artefacts procedure (SAS 1996). The MIXED procedure imple-
was tested using a priori comparisons (Underwood ments a generalisation of the standard linear model
1997). After significant tests, effects involving grazing, that allows for proper incorporation of random effects.
either as a main effect or in interaction, were split into Model parameter estimates were fitted using the
2 comparisons: Among Controls (i.e. PC vs C) and restricted maximum likelihood method (Litell et al.
Exclusion vs Controls. The former tested for experi- 1996). For further details on the MIXED procedure, see
mental artefacts, while the latter tested for effect of Litell et al. (1996) and SAS (1996).
limpet removal. In some cases, artefacts might exist There is a problem of confounding effects associated
(see ‘Results’) and therefore to estimate the effect of with comparisons of means of treatments started in
limpets we compared Exclusion and Procedural Con- different seasons. For instance, in January 1998, differ-
trol quadrats. Comparisons were not non-independent ences between quadrats established in summer and
and non-orthogonal and therefore the probability of those established in winter could be due to different
Type I error was increased. However, we kept α = 0.05 start seasons as much as to the different time during
for each comparison to avoid excessive Type II errors which the changes occurred (an average of 18 mo for
(Underwood 1997). Algal species were pooled in 2 summer quadrats and 12 mo for winter quadrats). On
groups: (1) Fucus spp. distinguishing between primary the other hand, if means are compared after a fixed
and secondary cover, and (2) all other macroalgae time from establishment of experimental quadrats (e.g.
pooled (only primary cover). For cover of algae, the 12 mo), then differences could be due to the fact that
analysed data were percentage of cover in individual quadrats were sampled at different times of the year
quadrats. For barnacles, because cover at any date is (summer for summer quadrats and winter for winter
dependent on cover at the beginning of the experi- quadrats). If in the first case significant seasonal effects
ment, the analysed data were proportions of change might be due to the time elapsed since the experi-
Arrontes et al.: Effects of grazing by limpets 121
mental manipulations, in the second, seasonal effects 0.05). Slope and roughness of experimental quadrats
could be due to differences in the presence or abun- did not differ among any combination of treatments (no
dance of species at different times of the year. As this significant effects in ANOVA, data not shown). While
problem is unavoidable with the present design, data orientations were not statistically analysed, no obvious
were analysed 12 mo after the establishment of the trend was observed for quadrats under any treatment.
quadrats and at the end of the experiment, and a In Campiello, relative abundances of limpets at the
cautious interpretation of the results was made. In start of the experiment were 81.2 and 18.8% (N = 1267)
other cases, analyses were performed with data at the for Patella depressa and P. vulgata respectively. The
time in which maximum abundance was recorded. mean size (SE) was 1.46 ± 0.012 cm (N = 1029) for
Temporal trends were deduced from visual inspection P. depressa and 1.53 ± 0.035 cm (N = 238) for P. vul-
of graphs. gata. In Artedo, relative abundances of limpets were
90.3 and 9.7% (N = 859) for P. depressa and P. vulgata
respectively, and the mean sizes were 1.51 ± 0.018 cm
RESULTS (N = 776) and 1.47 ± 0.076 cm (N = 83) respectively.
Data on the density of limpets in some quadrats before
Initial conditions manipulations were lost and thus, the initial density of
limpets was estimated from untouched controls at the
Initial cover of barnacles was very variable and var- time of the first sampling. Initial densities were 59.13 ±
ied between 41 and 96%. Significant differences 4.42 (mean ± SE, N = 24) for Campiello and 42.5 ± 3.59
existed among sites (Table 1A). The triple interaction for Artedo. ANOVA revealed that differences in mean
(Grazing × Locality × Season) was also significant. The density were significant only between sites within
height on the shore of experimental quadrats oscillated each date, locality and season (Table 1B).
between 1.17 and 2.64 m above LAT. Despite this wide
range of variation, only significant differences in tidal
height were detected among sites (Table 1A). ANOVA Effects of limpet removal
revealed no significant main effects or interactions,
with associated p always above 0.3. No correlation Fences failed to completely exclude limpets in
existed between the height on the shore of the experi- exclusion quadrats (Fig. 1A); however, the density of
mental quadrats and barnacle cover (r = 0.063, p > limpets was considerably lower in exclusion quadrats
Table 1. Analysis of differences in initial conditions of experimental quadrats. (A) Barnacle cover and tidal height. For cover of
barnacles, variances were homogeneous after arcsine transformation. No transformation was needed for height on the shore.
(B) Density of limpets in unfenced control quadrats. No transformation needed (ns = non-significant, *p < 0.05, ***p < 0.001).
Because the analyses were performed to identify gross differences in initial conditions, pooling was not necessary
Source of variation df Error term MS F p MS F p
(A) Barnacles and tidal height –––––––––– Barnacles ––––––––––– ––––––– Tidal height ––––––
(a) Grazing = G 2 (d) 0077.18 1.09 ns 0.033 0.62 ns
(b) Locality = L 1 (h) 1142.88 4.68 ns 0.030 0.07 ns
(c) Season = Se 1 (f) 0238.97 0.28 ns 0.021 0.05 ns
(d) G × L 2 (i) 0071.10 3.23 ns 0.054 0.99 ns
(e) G × Se 2 (g) 0010.47 0.09 ns 0.010 0.19 ns
(f) L × Se 1 (h) 0855.84 3.51 ns 0.427 0.96 ns
(g) G × L × Se 2 (i) 0117.87 5.36 * 0.054 0.99 ns
(h) Date = D(L × Se) 4 (j) 0243.99 1.11 ns 0.446 1.46 ns
(i) G × D (L × Se) 8 (k) 0021.98 0.29 ns 0.055 0.98 ns
(j) Site = Si[D(L × Se)] 8 (l) 0219.80 5.01 *** 0.306 4.65 ***
(k) G × Si[D(L × Se)] 160 (l) 0075.85 1.73 ns 0.056 0.85 ns
(l) Residual 960 043.88 0.066
(B) Limpets
(a) Locality = L 1 (d) 3316.69 2.59 ns
(b) Season = Se 1 (c) 0165.02 0.19 ns
(c) L × Se 1 (d) 0858.62 0.67 ns
(d) Date(L × Se) 4 (e) 1281.06 2.03 ns
(e) Site[D(L × Se)] 8 (f) 0630.85 3.01 *
(f) Residual 320 0209.50
122 Mar Ecol Prog Ser 277: 117–133, 2004
than in the controls. Most of the limpets found within cant interaction reflected differences in the pooled
the fences were of small size and their numbers fol- abundance in both types of controls.
lowed seasonal variations in abundance in control At the time of highest density (November 1997),
quadrats (data not shown). The number of limpets in trochids were more abundant in exclusion quadrats
exclusion quadrats exhibited an increased trend than in controls (Fig. 1B). On some sampling dates
towards the end of the experiment, possibly reflect- (summer months, data not shown) virtually no animals
ing damage to the fences and previously unnoticed were found in control quadrats, while they were still
recruitment. At the time of highest abundance of present in exclusion quadrats. Large differences in
limpets (May 1997), differences existed between ex- density also existed between sites. ANOVA was per-
clusion and procedural control quadrats, but also be- formed for data collected in November 1997 (Table 2).
tween both types of control quadrats (Fig. 1A). Aver- Apart from the significant site effects, the analysis
age density was higher in procedural controls than in revealed significant differences between exclusion
untouched control quadrats. Apparently, fences quadrats and controls. No significant differences ex-
attracted limpets. Large differences among sites are isted between the 2 types of controls, though a trend of
also evident in Fig. 1A. ANOVA revealed that the larger density in the procedural control quadrats as
effects of grazing and site were significant (Table 2). compared to untouched controls was observed. In fact,
Analyses performed 3, 6 and 12 mo after manipula- if ANOVA was exclusively performed with control
tions (data not shown) rendered similar results, quadrats (data not shown), a significant difference
though the interaction Grazing × Site was also signif- between the 2 types of controls existed. As in the case
icant. Because the number of limpets in exclusion for limpets, trochids appeared to be attracted by
quadrats was fairly constant across sites, this signifi- fences.
Exclusion
A. Limpets
Procedural Control
Control
80 125
Campiello Artedo
100
60
75
40
50
No. of animals per quadrat
20
25
0 0
Site 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2
Date D1 D2 D1 D2 D1 D2 D1 D2
Season Summer Winter Summer Winter
B. Trochids
30 60
Campiello Artedo
20 40
10 20
0 0
Site 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2
Date D1 D2 D1 D2 D1 D2 D1 D2
Season Summer Winter Summer Winter
Fig. 1. Mean abundance of (A) limpets in the 3 grazing treatments in May 1997 and (B) trochids in November 1997. Left panels
are data from all sites pooled (N = 48). Right panels are abundance at individual sites. In all cases, data are mean ± SE
Arrontes et al.: Effects of grazing by limpets 123
Table 2. Analysis of differences in density of limpets (May 97) and Trochids (Nov 97). Variances were homogeneous after square
root transformation. In both cases, the value for 1 control quadrat lost in Artedo was replaced by the mean of the remaining
2 quadrats of the same treatment and site, and 1 df subtracted from the residual (ns = non-significant, *p < 0.05, ***p < 0.001)
(a) Source of df Error Limpets Trochids
(a) variation term MS F p MS F p
(a) Grazing = G 2 364.01 77.091
Exclusion vs
procedural control 1 Pooled-2 609.34 360.2511 *** 94.151 35.221 ***
Among controls 1 Pooled-2 118.45 5.00 * 3.45 1.29 ns
(b) Locality = L 1 (h) 110.55 0.05 ns 5.13 0.83 ns
(c) Season = Se 1 (f) 110.24 0.01 ns 23.721 50.621 ns
(d) G × L 2 Pooled-1 111.22 0.38 ns 1.48 0.52 ns
(e) G × Se 2 Pooled-2 110.62 0.37 ns 2.23 0.83 ns
(f) L × Se 1 (h) 120.78 1.90 ns 0.47 0.08 ns
(g) G × L × Se 2 Pooled-1 112.38 1.42 ns 1.94 0.68 ns
(h) Date = D(L × Se) 4 (j) 110.95 3.71 ns 6.18 0.69 ns
(i) G × D(L × Se) 8 (k) 111.92 1.24 ns 3.54 1.42 ns
(j) Site = Si[D(L × Se)] 8 (l) 112.95 2.56 * 9.01 3.38 *
(k) G × Si[D(L × Se)] 16 (l) 111.55 1.34 ns 2.48 0.93 ns
(l) Residual 95 111.15 2.66
(l) Pooled-1 error term: (i) + (k) 24 1.67 2.83
(l) Pooled-2 error term: (d) + (g) + (i) + (k) 281 1.69 2.67
Effect of limpet removal on macroalgae concluded that departure from normality had little
effect. Early colonisation, estimated as increases in
The most abundant algal species in exclusion primary cover, varied between localities and seasons
quadrats were Fucus spp. One year after manipulation, (significant interaction in ANOVA for primary cover,
several quadrats exhibited a secondary cover above Table 3). Colonisation was very rapid in Artedo in sum-
95%. Although all specimens that could be safely iden- mer quadrats and similar in both seasons in Campiello,
tified were F. vesiculosus, many small specimens re- and in Artedo in winter (Fig. 2). No other effect was
mained unidentified and we chose to keep the generic significant. Twelve months after manipulation, ANOVA
denomination. Except for 1 site in Artedo, in which revealed that the only significant effect was site. As
Fucus colonised both types of control quadrats, mea- shown in Fig. 3, sites within the same time period,
surable cover of Fucus only appeared in exclusion locality and season exhibited large differences in aver-
quadrats (Fig. 2). Only these quadrats were analysed age cover (e.g. more than 80% and less than 10% for
for differences in cover. The analysis is simplified by sites on one winter date in Artedo). Although mean
dropping the main effect of grazing and its interactions secondary cover appeared to be higher in summer
from the model and testing for spatial and temporal than in winter quadrats (Fig. 2), no effect of season was
effects on the growth of Fucus in exclusion quadrats. detected. In January 1998, with quadrats being 18 and
There were problems with analyses of cover of Fucus. 12 mo old, the results of ANOVA were identical (data
Even after transformation of data, heterogeneity of not shown). In October 1998, angular transformation
variances and gross departures from normality existed rendered homogeneous variances and almost no bi-
for some of the sampling dates. Lack of normality was modal distribution. Again, ANOVA revealed that site
a consequence of obvious bimodality of data within the was the only significant effect. This result suggests that
exclusion treatment, with a fraction of the quadrats trends inferred from previous analyses may be close to
lacking measurable cover of Fucus and a fraction with reality. However, differences among sites were smaller
large secondary covers. The analyses were made any- (Fig. 3).
way and therefore their results must be used to suggest Colonisation by Fucus was influenced by the height
possible trends rather than to construct unambiguous on the shore of individual quadrats. More Fucus
conclusions. However, non-parametric analyses of appeared in lower than in higher quadrats (Fig. 4,
variance were performed where clear bimodal data Table 4). No effect of locality existed but summer
existed (DISTLM v.2, Anderson 2001, 2003, McArdle & quadrats were significantly affected by height.
Anderson 2001). F -ratios and associated probabilities Excluding Fucus spp., a total of 27 taxa of algae were
were almost identical to those obtained with para- identified in the experimental quadrats. Their relative
metric ANOVAs (data not shown) and therefore, it is abundance was estimated as the percentage of the
124 Mar Ecol Prog Ser 277: 117–133, 2004
20
Summer. Campiello Winter. Campiello
15
Fucus primary cover (%)
10
5
0
20
Summer. Artedo Winter. Artedo
15
10
5
0
J J A SOND J FMAMJ J A SOND J FMAMJ J A SON J J A SOND J FMAMJ J A SOND J FMAM J J A SON
1996 1997 1998 1996 1997 1998
Exclusion Procedural control Control
100
Summer. Campiello Winter. Campiello
75
Fucus secondary cover (%)
50
25
0
100
Summer. Artedo Winter. Artedo
75 Fig. 2. Fucus spp. Changes in
cover. Top panels: primary cover
50
in exclusion quadrats alone.
25 Bottom panels: secondary cover.
Exclusion quadrats were sam-
0 pled for secondary cover on 2
additional dates, March and
J J A SOND J FMAMJ J A SOND J FMAMJ J A SON J J A SOND J FMAMJ J A SOND J FMAMJ J A SON October 1998. In all cases, data
1996 1997 1998 1996 1997 1998 are mean ± SE (N = 12)
Table 3. Analysis of differences in Fucus cover. Primary cover was analysed 6 mo after manipulation. Variances were still het-
erogeneous (Cochran’s C = 0.36, 0.01 < p < 0.05) after arcsine transformation. For analyses of secondary cover, arcsine trans-
formation rendered homogeneous variances though gross departure from normality occurred 12 mo after manipulations. For
October 1998, variances were homogeneous and close to normality after arcsine transformation (ns = non-significant, *p < 0.05,
**p < 0.01). Pooling of non-significant error terms did not alter the significance of tests for locality, season and their interaction,
and is not presented
Source of df Error Primary Secondary
variation term 6 mo 12 mo Oct 98
MS F p MS F p MS F p
(a) Locality = L 1 (d) 355.32 7.35 ns 1406.53 0.49 ns 1666.74 0.53 ns
(b) Season = Se 1 (c) 123.12 0.23 ns 6203.44 2.93 ns 1152.32 6.25 ns
(c) L × Se 1 (d) 516.89 10.691 * 2119.39 2.54 ns 1124.39 0.02 ns
(d) Date(L × Se) 4 (e) 148.38 0.21 ns 1835.69 0.40 ns 1247.23 0.80 ns
(e) Site[D(L × Se)] 8 (f) 223.43 2.02 ns 2083.85 3.83 ** 1560.72 2.51 *
(f) Residual 32 110.63 1543.99 1621.53
Arrontes et al.: Effects of grazing by limpets 125
Artedo Campiello
Summer Summer
Winter Winter
100
Fucus (%)
80
60
40
20
0
1.00 1.50 2.00 2.50
Tidal height (m)
Fig. 4. Fucus spp. Relationship between secondary cover of
individual exclusion quadrats 12 mo after the start of the
experiment and height on the shore
ences; possibly due to the large site effect. The pattern
of variation in abundance was similar for summer and
Fig. 3. Fucus spp. Mean secondary cover at individual sites winter quadrats in Campiello (Fig. 6). Highest cover
12 mo after the start of the experiment and in October 1998 was observed 10–11 mo after manipulations. In addi-
(only exclusion quadrats). Data are mean ± SE (N = 3) tion to low cover, no clear pattern was observed in
summer quadrats in Artedo. It could be hypothesised
that rapid growth of Fucus in these quadrats prevented
number of interceptions of individual taxa in relation to growth of other algae.
the total number of interceptions of all algae (Fucus
excluded), in all quadrats, on all sampling dates. Most
of the identified taxa were recorded sporadically. The Effect of limpet removal on barnacles
soft encrusting species Ralfsia verrucosa and the blue-
green Rivularia bullata were the most abundant taxa There was an overall trend of barnacles to decrease
and altogether comprised more than 75% of all obser- with time (Fig. 7). The decrease, however, was more
vations. Groups with more than 1% of relative abun- marked in exclusion quadrats. Apparent differences
dance of algae are presented in Fig. 5. also existed between seasons (more barnacles being
Measurable algal growth almost exclusively occur- lost in exclusion summer quadrats) and between sites
red in exclusion quadrats (Fig. 6) and therefore, only within each date, locality and season. Data were
these were analysed for differences in algal cover. analysed 12 mo after manipulations and in January
On the dates of highest abundance (May
and October 1997 for summer and winter
quadrats in Campiello, and November Table 4. Analysis of the relationship between the proportion of Fucus cover
1997 for both seasons in Artedo), the aver- and height on the shore. Logistic regression using SAS GLIMMIX (SAS 1996)
age percentage of cover varied between
more than 40% in some sites to less than Test of effects df p Parameter
Likelihood ratio test (χ2) estimates (SE)
5% in others (Fig. 6). The analysis was per-
formed with data from these sampling Location (random) 0.60 1 > 0.750<
dates. ANOVA revealed that only the site
was significant (Table 5). Although im- F-value Summer:
portant differences appeared to exist be- Season (fixed) 7.71 1, 44 0.008 1.421 (1.813)
Tidal height (fixed) 5.52 1, 44 0.023 –2.449 (0.512)
tween summer quadrats in Campiello and Intercept 3.458 (1.813)
Artedo, ANOVA failed to detect differ-
126 Mar Ecol Prog Ser 277: 117–133, 2004
sa 1998. For this date, an additional analysis was per-
rru a
ve alfsi
co
60 formed with summer quadrats. After 12 mo, variances
R
were still heterogeneous after angular transformation
Relative abundance (%)
and the comment made for Fucus cover is applicable.
45
p.
Analyses for both 12 mo and January 1998 data re-
sp
vealed a significant main effect of grazing, with exclu-
ix
hr
s
a
es
ium inc sum
inn p.
an
i fi d
lot
sion quadrats losing more barnacles than procedural
lm tus
id
30
a p sp
sp rust
Ca
at
ho
um rtuo
he ella
a
lum
controls and no differences among controls (Table 6A).
p.
int
un rph
)
xa
ph m to
on s st
sp
p.
sp usil
o
ta
This main effect, however, is not interpretable due to
ho ella
m
de
m rp u
u
p
5
op
15
yll
Lit hyll
(1
p.
Ul ium
the significant interactions. The site and the more rele-
uin
Ne oca
er
rs
m
ali
m
p
do
t
he
ho
lid
En
Os
t
ra
va vant interaction of grazing with the site were also
as
Au
Ge
Ot
Ce
Lit
0 M significant. In some sites, but not in others, the removal
Fig. 5. Relative abundance of macroalgae (Fucus excluded) in
of limpets led to a larger reduction of barnacles in
all quadrats at all sampling dates (see text). Only taxa with a relation to procedural controls. Both analyses also re-
relative abundance above 1% are identified vealed that differences among controls existed. These
differences were not consistent across
sites and though there was a trend in
some sites for procedural controls to
lose more barnacles than untouched
controls, in others, the opposite or no
trend was observed. As for the density
of limpets, the analyses suggested that
some artefacts caused by fences might
have occurred. Significant effects of
locality existed for the 12 mo data, with
more barnacles being lost in Campiello
than in Artedo. For the January 1998
data, the interaction between grazing
and season was significant, more
barnacles being lost in exclusion than
in controls treatments in quadrats
established in summer, with a smaller
difference in winter quadrats (Fig. 7).
No significant interaction between
differences among controls with sea-
sons was detected. Loss of barnacles
was also evident if only summer
quadrats in January 1998 were consid-
ered (Fig. 7). ANOVA for these data
rendered similar significant results
(Table 6B), but no significant differ-
ences between controls were found.
Loss of barnacles due to the exclu-
sion treatment was correlated with
cover of Fucus of individual quadrats
(Fig. 8). The proportion of barnacle
cover lost 12 mo after manipulations
was positively related to cover by
Fucus (r = 0.523, df = 46, p < 0.01). Note
that an inverse relationship existed
Fig. 6. Top panel: changes in the primary cover of macroalgae (Fucus between height on the shore and the
excluded) during the experiment (N = 12). Arrows indicate dates at which percentage of cover of Fucus. Simi-
maximum cover was observed at each locality and season. These dates were
used for ANOVA. Bottom panel: mean abundance of macroalgae at individual
larly, loss of barnacles was also in-
sites on the sampling dates with maximum cover (N = 3). In all cases, data are versely related to height on the shore
mean ± SE of the quadrats (data not shown, r =
Arrontes et al.: Effects of grazing by limpets 127
Table 5. Analysis of differences in cover by macroalgae quadrats under each grazing treatment in which
(Fucus excluded). Only exclusion quadrats were analysed. whelks were observed at least once is compared,
Variances were homogeneous and data were not transformed
whelks occurred more in exclusion than in control
(ns = non-significant, ***p < 0.001). Pooling of non-significant
error terms did not alter the significance of tests for locality, quadrats (Table 7). No significant differences were
season and their interaction, and is not presented observed between controls. Data suggested a rela-
tionship between secondary cover of Fucus 18 mo
Source of df Error MS F p after manipulation and the accumulated number of
variation term whelks counted in individual exclusion quadrats
manipulated in summer (Fig. 9B). While at high
(a) Locality = L 1 (d) 1096.81 4.08 ns covers of Fucus whelks were either fairly common or
(b) Season = Se 1 (c) 1144.44 0.10 ns rare, at low covers whelks were always rare. There
(c) L × Se 1 (d) 1455.70 1.69 ns could also be a relationship between the accumulated
(d) Date(L × Se) 4 (e) 1269.10 0.24 ns
number of whelks in exclusion quadrats and the
(e) Site[D(L × Se)] 8 (f) 1121.82 12.701 ***
proportion of barnacle cover lost (Fig. 9A). Again,
(f) Residual 321 1188.33
while both a small or large proportion of cover could
–0.370, df = 46, p < 0.05). For summer
quadrats in January 1998 (18 mo old) the
relationships still held (Fucus, r = 0.712,
df = 22, p < 0.01; height, r = –0.418, df =
22, p < 0.05).
The loss of barnacle cover may also
be a consequence of another indirect
effect. Specimens of the whelk Thais
lapillus were occasionally found within
the experimental quadrats. Occurrence
was very irregular, whelks being ob-
served in a low proportion of the visits
to individual quadrats and, when pre-
sent, in variable numbers, ranging from
1 to 37 specimens in 1 procedural con-
trol (Table 7). Although this observation
precluded any formal statistical com-
parison of the effect of grazing mani-
pulation as undertaken above, some
exploratory description is possible. If
all animals counted in all quadrats at
all sampling dates under each grazing
treatment are pooled, a trend can be
observed (Fig. 9A, Table 7). More ani-
mals appeared in exclusion than in
control quadrats. In addition, if data
are pooled for 3 periods, 0 to 6, 7 to 12
and 13 to 18 mo after manipulations,
then it can be observed that differences
became relevant in late sampling dates
(Fig. 9A). If the number of individual
Fig. 7. Top panels: changes in barnacle cover
(N = 12). Bottom panels: losses in barnacle
cover at individual sites 12 and 18 mo after the
start of the experiment (N = 3). In all
cases, data are mean ± SE
128 Mar Ecol Prog Ser 277: 117–133, 2004
be lost at low numbers of whelks, when whelks were Table 7. Total number of whelks Thais lapillus recorded in
fairly common, the loss of barnacles was always experimental quadrats from June 1996 to January 1998, range
of abundance in individual quadrats and number of quadrats
above 50%.
in which whelks were recorded at least once (occurrence).
Chi-square tests for 2 × 2 contingency tables (df = 1) are for
differences in occurrence of whelks under different combina-
DISCUSSION tions of treatments. There were 48 exclusion quadrats, 48 pro-
cedural controls and 47 controls (1 control quadrat was lost)
(ns = non significant, ***p < 0.001)
The experimental results supported the hypotheses
tested. Changes in the assemblages of the exclusion
quadrats were greater than those in both types of con- Treatment Total number Range Occurrence
trols. Exclusion promoted macroalgal growth and
Exclusion 406 1–26 32
induced the loss of barnacle cover. The ability of lim-
Procedural control 80 1–37 14
pets to control algal growth and to influence the abun- Control 20 1–5 8
dance of other animal species is not at all surprising as
χ2 tests
there is overwhelming evidence for this worldwide
Exclusion vs controls 25.68***
(Underwood & Jernakoff 1981, see review by Hawkins Among controls 1.97 ns
& Hartnoll 1983). The relevant aspects are the scales
Table 6. Analyses of differences in the percentage of cover lost by barnacles. After arcsine transformation, variances were still
heterogeneous for 12 mo data. Variances were homogeneous after transformation for data in January 1998. One missing replicate
and 1 erroneous value were replaced by the mean of the remaining 2 replicates of the same treatment at the same sites and 2 df
subtracted from the residual (ns = non-significant, *p < 0.05, **p < 0.01, ***p < 0.001)
Source of variation df Error term MS F p MS F p
(A) All experimental quadrats ––––––––– 12 mo –––––––– –––––––– Jan 98 –––––––
(a) Grazing = G 2 3262.73 6436.31
Exclusion vs procedural control 1 Pooled-2 3269.84 8.28 ** 5923.25 10.681 **
Among controls 1 Pooled-2 1433.58 1.10 ns 1136.57 2.05 ns
(b) Locality = L 1 (h) 15107.961 16.141 * 4995.27 6.14 ns
(c) Season = Se 1 (f) 5096.00 8.56 ns 1470.70 0.61 ns
(d) G × L 2 Pooled-1 1459.49 1.12 ns 1166.98 0.27 ns
(e) G × Se 2 Pooled-2 1597.06 1.51 ns 2041.09
Exclusion vs procedural control × Se 1 Pooled-2 2816.24 5.08 *
Among controls × Se 1 Pooled-2 1118.31 0.03 ns
(f) L × Se 1 (h) 1595.61 0.64 ns 1774.63 0.95 ns
(g) G × L × Se 2 Pooled-1 1165.80 0.41 ns 1178.62 0.13 ns
(h) Date = D(L × Se) 4 (j) 1936.26 1.97 ns 1813.64 1.34 ns
(i) G × D(L × Se) 8 (k) 1139.61 0.26 ns 1319.31 0.41 ns
(j) Site = Si[D(L × Se)] 8 (l) 1474.55 2.12 * 1608.16 3.05 **
(k) G × Si[D(L × Se)] 161 1543.29 1780.31
Exclusion vs procedural control × [D(L × Se)] 8 (l) 1586.30 2.62 * 1107.75 5.56 ***
Among controls × Si[D(L × Se)] 8 (l) 1803.17 3.59 ** 1715.40 3.59 **
(l) Residual 941 1223.55 1199.20
Pooled-1 error term: (i) + (k) 241 1408.76 1626.64
Pooled-2 error term: (d) + (g) + (i) + (k) 281 1395.03 1554.66
(B) Summer quadrats in January 1998 (18 mo)
(a) Grazing = G 2 7763.861
Exclusion vs procedural control 1 Pooled-2 8454.021 12.211 **
Among controls 1 Pooled-2 721.70 1.04 ns
(b) Locality = L 1 (d) 917.85 1.71 ns
(c) G × L 2 Pooled-1 189.02 0.24 ns
(d) Date = D(L) 2 (f) 538.15 0.65 ns
(e) G × D(L) 4 (g) 491.35 0.54 ns
(f) Site = Si[D(L)] 4 (h) 822.24 3.15 *
(g) G × Si[D(L)] 8 918.36
Exclusion vs procedural control × Si[D(L)] 4 (h) 1406.931 5.38 **
Among controls × Si[D(L)] 4 (h) 485.49 1.86 ns
(h) Residual 46 261.31
Pooled-1 error term: (e) + (g) 12 776.02
Pooled-2 error term: (c) + (e) + (g) 14 692.17
Arrontes et al.: Effects of grazing by limpets 129
of variability of this effect, the nature of the A
interactions behind the measurable effects and Exclusion
200
Procedural control
Total no. of whelks
how this variability and interactions change
over large spatial scales. Control
150
An identical experimental design was used to
evaluate the effects of grazing on the structure 100
of moderately exposed rocky shores on the west
coast of Sweden (Lindegarth et al. 2001). How- 50
ever, a formal comparison of the results from
0
both experiments, although possible, is not 0-6 7-12 13-18
useful. Differences in the composition of the Period (months)
intertidal assemblages and contrasting environ-
mental conditions preclude any sensible inter- B 18 months Campiello Artedo
pretation of a common statistical analysis since
Proporion of barnacle
60 1.0
any observed difference might be a conse-
No. of whelks
quence of multiple factors. The shores of the 0.8
cover lost
north of Spain and the west of Sweden differ in 40
0.6
the main key grazer: limpets in Spain and Litto-
0.4
rina littorea in Sweden (Cervin & Åberg 1997, 20
0.2
Viejo et al. 1999). Dominant algal groups in
Sweden include filamentous and crustose red 0 0.0
algae and ephemeral greens (Lindegarth et al. 0 20 40 60 80 100 0 20 40 60
2001), while virtually no macroalgae are pre- Fucus (%) No. of whelks
sent at mid-tidal levels in northern Spain
Fig. 9. (A) Total number of whelks counted in different periods during
(Anadón 1983). In addition, obvious different the experiment. For the initial period (0 to 6 mo) data are accumulated
thermal and radiation regimes exist in both numbers from 8 sampling dates. For the periods 7 to 12 and 13 to
places, with ice-scouring being a fairly common 18 mo, data are accumulated from 3 sampling dates. The period 13
event in Sweden. Tides in northern Spain are to 18 mo is for quadrats manipulated in summer. (B) Relationship
between the accumulated number of whelks and the percentage of
cover of Fucus (left) and between the proportion of barnacle loss and
Campiello Artedo accumulated number of whelks (right). In both cases, data are from
summer individual quadrats after 18 mo of experimental manipulation
Summer Summer
Winter Winter
12 months semi-diurnal with a maximum range of 4.3 m, while in
0.8 Sweden the tidal range is very narrow and often
0.6 dwarfed by unpredictable sea-level variations (Johan-
Proportion of barnacle cover lost
0.4 nesson 1989). However, qualitative comparisons are
0.2 possible. Site effects were important in both studies.
0.0
The role of grazing in the change of the structure of
assemblages across dates and sites was weak and
-0.2
inconsistent in Sweden, and it is concluded by Linde-
garth et al. (2001) that grazing is less important there
18 months than in semi-exposed shores in other parts of the
1.0
world. Our results support this conclusion as we show
0.8 that grazing affected the structure of assemblages by
0.6 limiting the development of ephemerals and algal
0.4
canopies.
Other studies of the grazing activity of limpets over-
0.2
lap with the experiment reported in this paper. Using
0.0 wax discs placed on the rock surface in Campiello and
0 20 40 60 80 100 Artedo, Jenkins et al. (2001) estimated that an average
Fucus (%) of 26% of rock surface was scraped by limpets at
Fig. 8. Relationship between the proportion of barnacle cover
natural densities in 2 wk, without any seasonal trend
lost and cover of Fucus in individual exclusion quadrats 12 and with similar values for both localities. This amount
and 18 mo after the start of the experiment may appear low and, in fact, it is significantly lower
130 Mar Ecol Prog Ser 277: 117–133, 2004
than estimated rates for southern England or southern exclusion quadrats reverted to a situation similar to
Portugal (Jenkins et al. 2001). However, if we consider that before the manipulations. The persistence of
that an escape size for Fucus is only attained several dense cover of macroalgae depends upon the con-
months after the settlement of zygotes, then it may be tinuity of the manipulation and thus the grazer- and
enough to prevent the development of a Fucus canopy. barnacle-dominated assemblage and the algal patches
In addition, grazing by limpets is not a random process cannot be considered alternate stages at these tidal
(e.g. Chelazzi et al. 1998b) and the spatial organisation levels (Petraitis & Dudgeon 1999).
of foraging might be related to the availability of food. The second prediction, decreased barnacle cover
The estimated 26% of rock surface grazed in Jenkins after limpet removal, was also supported by the exper-
et al. (2001) is a mean value but individual wax discs imental results. However, the effect was measurable
presented large variability ranging from 0 to virtually only after 12 mo from the beginning of the manipula-
100% of the surface scraped. If the density of limpets tions. The lag in the effect is interpreted as only when
were well below natural levels, the grazing pressure algal cover was important and persisted for some
might not be enough to prevent algal growth, in which months did barnacles start to decline. Another relevant
case the probability of escape of individual algal pro- result is that the effect of the grazing treatment is site-
pagules would increase. No estimates on the efficiency dependent and presumably reflects site effects in the
of scraping in removing algal propagules was avail- growth of Fucus and other macroalgae. Results were
able, but reports on the effects of limpets on the consistent across seasons, dates and localities. The
microflora growing on the rocks in Australia (Under- most recurrent result was that small-scale hetero-
wood & Jernakoff 1981) or studies on the anatomical geneity was important (significant site effects). Site
structure of the radula (Steneck & Watling 1982) effects might reflect environmental differences among
suggest that scraping may be very efficient. Radula sites. Before manipulations, sites exhibited differences
strokes virtually collect all algae from the rock surface. in barnacle cover and abundance of limpets. In addi-
The experimental design allowed us to test the con- tion, they had different mean tidal heights but were
sistency of the effect of grazing spatially and over time. similar in orientation, slope and roughness. Other fac-
No effect of start season or date was detected, as the tors, which could potentially influence the composition
interaction of grazing with these 2 factors was usually and dynamics of the assemblage, such as the prevail-
not significant. This is possibly due to the low variabil- ing direction of waves or exposure were not investi-
ity in pressure grazing by limpets over time (Jenkins gated. Site effects may be generated by multiple fac-
et al. 2001). However, the season of removal had a tors acting independently or synergistically. Tidal
significant effect on macroalgal growth at 1 locality, height alone may influence the growth rate of algal
resulting in a more rapid colonisation in Artedo during propagules (Raffaelli & Hawkins 1996). The direction
summer than in winter. This may be related to the of incoming waves and the distance to sources of
availability of propagules after the removal of limpets, propagules may affect not only the rate of develop-
in a similar mode as season for scraping influenced ment of algal canopies but also the possibility that
colonisation of experimental patches by Fucus in some algae (e.g. Fucus) could even reach individual
British Columbia (Kim & DeWreede 1996). Twelve and experimental quadrats (Arrontes 2002).
18 mo after manipulation, however, the significant There is another effect related to limpet removal: in-
interaction vanished. In part, this could be due to the creased densities of trochids (mainly Gibbula umbili-
method of estimation of macroalgal abundance. By calis). In this case, a clear interpretation is not possible.
estimating the percentage of cover of macroalgae, The increased density might be an indirect effect
differences in abundance under high and low recruit- caused either by the increase of food on the rock sur-
ment are only detected at early stages. After some face or by the growth of macroalgae, but no relation-
time, as individual plants grow, secondary cover may ship between abundance of algae and trochids was
be similar even with very different density of Fucus found. Alternatively, this increase in trochid density
plants. However, site effects were significant and might be an artefact. Grazers have been observed to be
surely reflect differences among sites (such as tidal attracted by fences and open cages in previous studies
height) independently of experimental manipulations. (e.g. Underwood 1980). A similar trend of increased
Apart from Fucus, other algae were abundant in some density in procedural controls was also observed in
quadrats, mainly Ralfsia-like soft encrusting species limpets and whelks. Some evidence of artefacts also
and ephemerals. In general, their abundance de- appeared in barnacle cover. Potential artefacts were
creased as the Fucus canopy developed. not always statistically significant, but in most cases
After the last sampling date in October 1998, all differences between both types of controls were always
fences were removed. Limpets quickly reinvaded the in the same direction. Significant effects of procedural
experimental quadrats and after 3 yr, all identifiable controls, even with clear exclusion effects, may gener-
Arrontes et al.: Effects of grazing by limpets 131
ate results that are not logically interpretable (Under- nipulation of abundance (e.g. presence/absence) of
wood 1997). In general, however, and due to the mag- the species assumed to exert the indirect effect and
nitude of the change in exclusion quadrats vs proce- (2) manipulation of the abundance of the third species
dural controls, we do not think that artefacts were directly responsible for changes in the abundance of
strong enough to invalidate the main conclusions. the focal species. Additional treatments should control
Both direct and indirect effects are responsible for for artefacts associated with manipulations. To con-
the observed results. The increase of algae is surely an clude that a decrease in barnacle cover is not a direct
immediate consequence of the cessation of limpet consequence of limpet removal but of the growth
grazing and is interpreted as a direct effect. Several of macroalgae, 2 additional experimental treatments
lines of evidence, however, suggest indirect effects on should consider the growth of macroalgae while
barnacles: (1) Limpets have been shown to remove limpets were still present, and the removal of limpets
juvenile and larvae of barnacles, either by accidental with no growth of algae. No experiments of this kind
or deliberate ingestion or by dislodging specimens were performed in this work to identify indirect effects
while crawling over the barnacles (e.g. Wootton 1993). and it is difficult to see how these treatments could be
If a measurable direct effect on barnacles existed it set up. However, considering the reported evidence
should be negative and thus removal of limpets should based on correlations and life-cycles, we are confident
promote an increase in barnacle cover. (2) Towards the that indirect effects exist.
end of the experiment, the rate of loss of barnacles was Our results are summarised in Fig. 10. We present
positively related to algal cover in exclusion quadrats. averaged species interactions but in some sites or
The direct causes might include overgrowth of barna- quadrats the picture may be completely different, as in
cles by algae, reduced water flow limiting availability some quadrats no growth of macroalgae occurred and
of food and inhibition of recruitment. (3) Sporadically, barnacle cover remained unchanged. Of course, there
juvenile whelks were found among algae in exclusion are more links in the assemblage than those in Fig. 10.
quadrats more often than in controls. The latter effect For instance, limpets have been shown to have a small
was also observed in previous studies such as the mon- deleterious effect on barnacles, and whelks are often
itoring program of the effects of the Torrey Canyon oil observed to feed on juvenile limpets (B. Martínez, pers.
spill (summarised in Raffaelli & Hawkins 1996). Juve- obs.). In addition, the interactions between canopy Fu-
nile whelks in Campiello and Artedo tend to aggregate cus and limpets may be complex and generate cycles
among algae, crevices or underneath boulders, partic- (Hawkins & Hartnoll 1983, Burrows & Hawkins 1998).
ularly on clear sunny days (B. Martínez, unpubl.). In The graph is also incomplete because other groups are
short, the indirect effect of limpets on barnacles should ignored; the complete assemblage should include
be positive and mediated by the inhibition of algal small littorinids, acari, insects, grapsid crabs, and inter-
growth. Similarly, the effects of algae on barnacles spersed Lychina pygmaea and Mytilus patches with
may be complex and include additional indirect effects their associated fauna. Fig. 10 only represents a subset
mediated by increasing abundance of juvenile whelks. of a more complex assemblage and a conspicuous
Indirect effects appear to be common in intertidal spatial heterogeneity. Finally, the graph would only be
systems. Menge (1995) estimated that 40 to 50% of the complete if the pattern of arrival of algal propagules
changes occurring after perturbations might be due to and larvae could be incorporated. It is evident that
indirect effects. Indirect effects include a great variety
of species interactions but all of them can be grouped
( -)
into 9 types (Menge 1995). Habitat facilitation is one of Limpets
these types and occurs when one species improves the Trochids
(- )
habitat of another species by altering the abundance of ( -)
a third species (Fairweather 1990). Habitat facilitation (+)
(+)
appears to be the type of indirect effect of limpets on Algae
barnacles described in this work. There are many
examples of indirect effects in intertidal communities ( -) ( -) Whelks
(see Menge 1995 for references). For example, in the (- )
San Juan Islands (Washington), removal of macroalgae Barnacles
by chitons enhances the growth of microalgae, which
are the food for acmaeid limpets (Dethier & Duggins Fig. 10. Summary of species interactions at mid-tidal levels in
1984). northern Spain suggested by various evidence in this paper.
Only groups and links investigated in this paper are pre-
Unambiguous identification of indirect effects de-
sented. Continuous lines: direct effects; discontinuous lines:
mands proper experimental designs involving at least indirect effects; grey line: doubtful effect. The thickness of the
2 factors in a crossed design (Anderson 1999): (1) ma- lines relates to intensity of the effect
132 Mar Ecol Prog Ser 277: 117–133, 2004
much additional work is needed for the unambiguous European Marine Science and Technology Conference,
identification and quantification of direct and indirect Lisbon, 23–27 May 1998: Project synopses, Vol 1: Marine-
systems. European Commission DG 12, Luxembourg,
effects, for the identification of the causes of the site
p 361–376
effects and to discover the role of the species ignored Chelazzi G, Santini G, Della Santina P (1998b) Route selection
in this work. in the foraging of Patella vulgata (Mollusca: Gastropoda).
J Mar Biol Assoc UK 78:1223–1232
Dayton PK, Tegner MJ, Parnell PE, Edwards PB (1992) Tem-
Acknowledgements. The experiment was funded by the poral and spatial patterns of disturbance and recovery in a
European Union as a part of the EUROROCK-project, MAS3- kelp forest community. Ecol Monogr 62:421–445
CT95-0012. Dethier MN, Duggins DO (1984) An ‘indirect commensalism’
between marine herbivores and the importance of com-
petitive hierarchies. Am Nat 124:205–219
LITERATURE CITED Dethier MN, Duggins DO (1988) Variation in strong interac-
tions in the intertidal zone along a geographical gradient:
Anadón R (1983) Zonación en la costa asturiana: variación a Washington-Alaska comparison. Mar Ecol Prog Ser 50:
longitudinal de las comunidades de macrófitos en difer- 97–105
entes niveles de marea. Invest Pesq 47:125–141 Fairweather PG (1990) Is predation capable of interacting
Anadón R, Niell FX (1981) Distribución longitudinal de with other community processes on rocky reefs? Aust
macrófitos en la costa asturiana (N de España). Invest Pesq J Ecol 15:453–464
45:143–156 Farrell TM (1991) Models and mechanisms of succession: an
Anderson MJ (1999) Distinguishing direct from indirect example from a rocky intertidal community. Ecol Monogr
effects of grazers in intertidal estuarine assemblages. 61:95–113
J Exp Mar Biol Ecol 234:199–218 Foster MS (1990) Organization and macroalgal assemblages
Anderson MJ (2001) A new method for non-parametric multi- in the Northeast Pacific: the assumption of homogeneity
variate analysis of variance. Aust Ecol 26:32–46 and the illusion of generality. Hydrobiologia 192:21–33
Anderson MJ (2003) DISTLM v.2: a FORTRAN computer pro- Hawkins SJ (1981) The influence of season and barnacles on
gram to calculate a distance-based multivariate analysis the algal colonization of Patella vulgata exclusion areas.
for a linear model. Department of Statistics, University of J Mar Biol Assoc UK 61:1–15
Auckland, New Zealand Hawkins SJ, Hartnoll RG (1983) Grazing of intertidal algae by
Anderson MJ, Underwood AJ (1997) Effects of gastropod marine invertebrates. Oceanogr. Mar Biol Annu Rev 21:
grazers on recruitment and succession of an estuarine 195–282
assemblage: a multivariate and univariate approach. Hawkins SJ, Harnoll RG, Kain JM, Norton TA (1992) Plant-
Oecologia 109:442–453 animal interactions on hard substrata in the north-east
Arrontes J (1993) Nature of the distributional boundary of Atlantic. In: John DM, Hawkins SJ, Price JH (eds) Plant-
Fucus serratus on the north shore of Spain. Mar Ecol Prog animal interactions in the marine benthos. Clarendon
Ser 93:183–193 Press, Oxford, p 1–32
Arrontes J (2002) Mechanisms of range expansion in the Jenkins SR, Arenas F, Arrontes J, Bussell J and 10 others
intertidal brown alga Fucus serratus in northern Spain. (2001) European-scale analysis of seasonal variability in
Mar Biol 141:1059–1067 limpet grazing activity and microalgal abundance. Mar
Benedetti-Cecchi L, Cinelli F (1994) Recovery of patches in an Ecol Prog Ser 211:193–203
assemblage of geniculate coralline algae: variability at Johannesson K (1989) The bare zone of Swedish rocky shores:
different successional stages. Mar Ecol Prog Ser 110:9–18 why is it there? Oikos 54:77–86
Benedetti-Cecchi L, Cinelli F (1997) Confounding in field Johnson LE (1992) Potential and peril of field experimenta-
experiments: direct and indirect effects of artefacts due to tion: the use of copper to manipulate molluscan herbi-
the manipulation of limpets and macroalgae. J Exp Mar vores. J Exp Mar Biol Ecol 160:251–262
Biol Ecol 209:171–184 Kennelly SJ (1987) Physical disturbances in an Australian
Benedetti-Cecchi L, Bulleri F, Cinelli F (2000) The interplay of kelp community. I. Temporal effects. Mar Ecol Prog Ser 40:
physical and biological factors in maintaining mid-shore 145–153
and low-shore assemblages on rocky coasts in the north- Kim JH, DeWreede RE (1996) Effects of size and season of
west Mediterranean. Oecologia 123:406–417 disturbance on algal patch recovery in a rocky intertidal
Boaventura D, Alexander M, Della Santina P, Smith ND, Ré P, community. Mar Ecol Prog Ser 133:217–228
Cancela da Fonseca L, Hawkins SJ (2002) The effects of Lindegarth M, Åberg PA, Cervin G, Nilsson PG (2001) Effects
grazing on the distribution and composition of low-shore al- of grazing on the structure of mid-shore, intertidal as-
gal communities on the central coast of Portugal and on the semblages on moderately exposed rocky shores on the
southern coast of Britain. J Exp Mar Biol Ecol 267:185–206 Swedish west coast. Mar Ecol Prog Ser 212:29–38
Burrows MT, Hawkins SJ (1998) Modelling patch dynamics Litell RC, Milliken GA, Striuo WW, Wolfinger RD (1996) SAS
on rocky shores using deterministic cellular automata. system for mixed models. SAS Institute, Cary, NC
Mar Ecol Prog Ser 167:1–13 McArdle BH, Anderson MJ (2001) Fitting multivariate models
Cervin G, Åberg P (1997) Do littorinids affect the survival of to community data: a comment on distance-based redun-
Ascophyllum nodosum germlings? J Exp Mar Biol Ecol dancy analysis. Ecology 82:290–297
218:35–47 Menge BA (1995) Indirect effects in marine rocky intertidal
Chelazzi G, Hawkins SJ, Åberg P, Arrontes J, Myers A, Paula interaction webs: patterns and importance. Ecol Monogr
J (1998a) Interactions of physical and biological factors 65:21–74
in the surf and swash zone of European rocky shores. Petraitis PS, Dudgeon SR (1999) Experimental evidence for
In: Barthel KG, Barth H, Bohle-Carbonell M, Fragakis C, the origin of alternative communities on rocky intertidal
Lipiatou E, Martin P, Ollier G, Weydert M (eds) Third shores. Oikos 84:239–245
Arrontes et al.: Effects of grazing by limpets 133
Raffaelli D, Hawkins S (1996) Intertidal ecology. Chapman & Underwood AJ (1997) Experiments in ecology. Their logical
Hall, London design and interpretation using analysis of variance. Cam-
Robles CD, Cubit J (1981) Influence of biotic factors in an bridge University Press, Cambridge
upper intertidal community: dipteran larvae grazing on Underwood AJ, Jernakoff P (1981) Effects of interactions
algae. Ecology 62:1536–1547 between algae and grazing gastropods on the structure of
SAS (1996) SAS/SYSTAT software: changes and enhance- a low-shore intertidal algal community. Oecologia (Berlin)
ments through release 6.11. SAS Institute, Cary, NC 48:221–233
Steneck RS, Watling L (1982) Feeding capabilities and limita- Underwood AJ, Petraitis PS (1993) Structure of intertidal
tion of herbivore molluscs: a functional group approach. assemblages in different locations: how can local pro-
Mar Biol 68:299–319 cesses be compared? In: Ricklefs R, Schluter D (eds) Spe-
Stephenson TA, Stephenson A (1972) Life between tidemarks cies diversity in ecological communities. University of
on rocky shores. WH Freeman, San Francisco Chicago Press, Chicago, p 38–51
Underwood AJ (1978) An experimental evaluation of compe- Viejo RM, Åberg P, Cervin G, Lindegarth M (1999) The inter-
tition between three species of intertidal gastropods. active effects of adult canopy, germling density and graz-
Oecologia (Berl) 33:185–202 ing on germling survival of the rockweed Ascophyllum
Underwood AJ (1980) The effects of grazing by gastropods nodosum. Mar Ecol Prog Ser 187:113–120
and physical factors on the upper limits of distribution of Winer BJ (1971) Statistical principles in experimental design,
intertidal macroalgae. Oecologia (Berl) 46:201–213 2nd edn. McGraw-Hill, New York
Underwood AJ (1984) Vertical and seasonal patterns in com- Wootton JT (1993) Indirect effects and habitat use in an inter-
petition for microalgae between intertidal gastropods. tidal community: interaction chains and interaction modi-
Oecologia (Berl) 64:211–222 fications. Am Nat 141:71–89
Editorial responsibility: Antony Underwood (Contributing Submitted: April 29, 2003; Accepted: April 21, 2004
Editor), Sydney, Australia Proofs received from author(s): August 6, 2004
Vol. 277: 117–133, 2004 Published August 16
Mar Ecol Prog Ser
Effect of grazing by limpets on mid-shore species
assemblages in northern Spain
J. Arrontes1,*, F. Arenas2, C. Fernández1, J. M. Rico1, J. Oliveros1, B. Martínez3,
R. M. Viejo3, D. Alvarez1
1
Departamento de Biología de Organismos y Sistemas, Universidad de Oviedo, 33071 Oviedo, Spain
2
Marine Biological Association of the United Kingdom, Prospect Place, Plymouth PL1 2PB, UK
3
Área de Biodiversidad y Conservación, Escuela Superior de Ciencias Experimentales y Tecnología,
Universidad Rey Juan Carlos, 28933 Móstoles, Madrid, Spain
ABSTRACT: Limpets were excluded from barnacle-dominated areas in 2 semi-exposed localities on
the shore of northern Spain. We tested the hypotheses that a macroalgal canopy would develop and
barnacle cover would decrease in exclusion quadrats. Spatial replication considered localities and
sites within localities, while temporal replication considered 2 seasons of removal and 2 start dates
within each locality and season. The experimental methods included exclusion quadrats (using
fences), procedural controls (half fences) and untouched controls. Results supported both hypotheses.
Limpet removal caused exclusion quadrats to change over time more than controls. An algal canopy
dominated by Fucus vesiculosus developed and more barnacles were lost in exclusion quadrats than
in controls. In addition, more trochids (largely Gibbula spp.) and whelks (Thais lapillus) were found
in exclusion quadrats. Site effects were significant for virtually all variables analysed but no effect of
locality, season or date was found. An indirect effect mediated by the development of an algal canopy
appeared to be responsible for the loss of barnacles in exclusion quadrats. This indirect effect might
have been reinforced by an additional indirect effect of algae mediated by increased densities of
whelks under the algal canopy. Altogether, our results agree with those in semi-exposed shores in
other parts of the world but not with results obtained in the west coast of Sweden, which has
contrasting environmental conditions and a different dominant grazer species.
KEY WORDS: Grazing · Community structure · Indirect effects · Patella spp. · Fucus spp. · Spain
Resale or republication not permitted without written consent of the publisher
INTRODUCTION of the grazers, the relative magnitudes of the feeding
rate and the growth rate of the algae (Underwood &
It is well documented that the composition and dy- Jernakoff 1981). At lower tidal levels or in areas with
namics of species assemblages in mid- and high tidal increased shading or wetness conditions, the relative
levels in rocky shores are strongly influenced by the importance of grazing as a structuring force may de-
activity of grazers (Hawkins & Hartnoll 1983). The crease as the growth rates of algae increase. Seasonal
most evident effect of grazing at the high shore is the and spatial variation in the intensity of recruitment is
removal of macroscopic algal species leaving the pri- an additional source of variability of the outcome of the
mary substratum occupied by sessile animals; how- interaction. In experimental shore ecology, there is
ever, the assessment of the effect of grazers is far from ample evidence that timing of manipulations can sig-
being that simple. Even if we are limited to the direct nificantly influence the rate of recovery and the spe-
removal of algae, there is an interaction between graz- cies sequence. Examples abound and cover a great
ing and the effect of physical factors (Underwood variety of habitats such as limpet dominated areas
1980). The final effect depends on the mode of feeding (Hawkins 1981), subtidal kelp forests (Kennelly 1987,
*Email: arrontes@uniovi.es © Inter-Research 2004 · www.int-res.com
118 Mar Ecol Prog Ser 277: 117–133, 2004
Dayton et al. 1992), assemblages of geniculate coral- replication at different time scales. Recent examples of
line algae (Benedetti-Cecchi & Cinelli 1994) or fucoid studies with adequate spatial replication can be found
dominated beds (Kim & DeWreede 1996). The effect of in Benedetti-Cecchi et al. (2000) and Boaventura et
grazers is not, however, limited to the removal of algae al. (2002), the former also with proper temporal
and very often grazers can affect other animal species replication.
through competitive interactions (Underwood 1978, It is obvious that not all species in the grazer assem-
Dethier & Duggins 1984). In this case, care must be blage have the same effect on algae (Underwood &
taken when generalising because the intensity of com- Jernakoff 1981, Underwood 1984). Often, this is the
petition may vary spatially and temporally in response consequence of a single species or a reduced group of
to the densities and identities of competitors and avail- species having a disproportionate effect. Limpets are
ability of food (Underwood 1984). Grazers may also very often the key grazers in mid- and high tidal
affect the rate of succession (e.g. Farrel 1991) or cause levels (Hawkins et al. 1992). In different parts of the
different assemblages to develop (e.g. Anderson & world, other gastropods (Cervin & Åberg 1997, Viejo
Underwood 1997). et al. 1999), chitons (Dethier &Duggins 1984) or even
Apart from direct effects, indirect effects of grazers insect larvae (Robles & Cubit 1981) may play relevant
are common and have been profusely described in roles.
shore environments (e.g. Menge 1995). These indi- The aim of this work was to assess the role of domi-
rect effects can usually be grouped into several nant grazers (limpets) on the structure of species as-
types, one of which is habitat facilitation: the activity semblages in mid-tidal levels of northern Spain at dif-
of one species promoting habitat transformation, ferent spatial and temporal scales. Barnacles are the
which makes it suitable for colonisation by other main space occupiers, in some zones covering up to
species. In some cases, there is a dual effect of the more than 90% of available rock surface. Virtually no
key grazer on a third species, a direct deleterious macroalgae are present and the grazer assemblage is
effect through exploitative or interference competi- limited to a small number of species (Anadón 1983).
tion and an indirect positive effect mediated by habi- Structurally, the system is similar to those found in
tat facilitation (Menge 1995). other parts of the world at the same tidal level
The structuring processes may vary between distant (Stephenson & Stephenson 1972, Raffaelli & Hawkins
areas and, in fact, some studies report an important 1996). The experimental design was intended to make
variability among different geographic zones even broad comparisons of the role of grazers in high and
when they have equivalent species composition (Dethier mid-tidal levels on European shores. The experimental
& Duggins 1988, Boaventura et al. 2002). Obviously, design considered temporal and spatial variability at
broad-scale comparisons of the effect of grazers are 2 different scales: among and within seasons, and
unavoidable in seeking generality of structuring forces among and within localities. Within the framework of
in shore communities. However, making generalisa- an international project (Chelazzi et al. 1998a), identi-
tions from particular studies performed in different cal studies were performed on other European shores
parts of the world at different times may be erroneous with the same experimental design at the same dates:
due to spatial and temporal variability (Foster 1990, Sweden (Lindegarth et al. 2001), Isle of Man, southern
Underwood & Petraitis 1993). Comparisons are often England and southern Portugal. Similar species assem-
difficult due to variable experimental designs and dif- blages but contrasting environmental conditions were
ferent degrees of spatial replication. Even in the case identified in different locations, such as the Isle of
of equivalent designs, variability in time and space Man, southern England and northern Spain. In other
may preclude any comparison (including qualitative cases, both environmental conditions and species dif-
comparisons) of studies performed at single distant fered (e.g. Sweden and northern Spain).
localities at different times. Different localities within
the same shore, although only a few kilometres apart,
may differ, and usually they do, in average orientation, MATERIALS AND METHODS
slope, wave exposure and intensity of recruitment.
Consequently, both density and composition of the Localities and species. The study was carried out
grazer assemblage as well as abundance and growth from May 1996 to October 1998 in 2 localities on the
rates of algae may differ among close locations; hence north shore of Spain, Campiello (43° 33’ N, 6° 24’ W)
the relative importance of grazing as a structuring and Artedo (43° 34’ N, 6° 12’ W). Both localities are
agent. The practical implication is that to cope with moderately exposed to wave action and have a gently
this expected variability, any large scale comparison sloping rock platform with many boulders. The rock
among shores must include replication at different platform in Campiello faces north and is slightly more
spatial scales, from metres to kilometres, and temporal exposed than in Artedo, which faces east. The nature
Arrontes et al.: Effects of grazing by limpets 119
of the rock is the same in both localities, largely replicated at different sites. At the lower part of the
quartzite with some veins of slate. The maximum tidal barnacle-dominated zone, 8 experimental sites inter-
range of spring tides is around 4.3 m. Intertidal com- spersed along the intertidal platforms in Artedo and
munities are very similar in the 2 localities and have 8 in Campiello were selected in late May 1996, before
been described elsewhere (Anadón & Niell 1981, the start of the experimental manipulations. Sites were
Anadón 1983, Arrontes 1993). Briefly, the high shore around 15 m long, at least 30 m apart and had enough
between roughly 3 and 1.5 m above Lowest Astronom- positions to install nine 50 × 50 cm experimental
ical Tide (LAT) is dominated by barnacles and grazing quadrats (see below). Criteria for the selection of the
gastropods. Primary space is largely occupied by 2 quadrats were (1) an inclination between 0 and 45°,
species of barnacles, Chthamalus montagui and C. (2) simple topography and absence of large cracks,
stellatus, small patches of the lichen Lichyna pygmaea crevices or pools, (3) cover of barnacles (Chthamalus
and ephemeral blue green algae. Two species of spp.) above 40%, (4) being at least 1 m apart from the
limpets, Patella depressa and P. vulgata, coexist in this nearest quadrat and (5) absence of perennial macroal-
zone. The relative abundance of both species varies in gae and less than 5% cover of ephemerals. No distinc-
the tidal range and along the shore. Other grazers are tion between stable bedrock and large boulders was
trochids (Gibbula spp. and Monodonta lineata) and made. Temporal variability also considered 2 scales,
small littorinids (species of the Littorina saxatilis group start season (summer and winter) and 2 start dates
and Melaraphe neritoides). In both localities, the within each season, which were different for each
predatory whelk Thais (= Nucella) lapillus is present at locality. Summer start dates were each of the 4 con-
low densities. Below the grazer-dominated zone, secutive spring tides in June–July 1996. Two start
between 1.5 and 0.75 m above LAT, fucoids monopo- dates were randomly assigned to each locality. Winter
lise the space (Fucus vesiculosus and F. serratus in start dates were the 4 spring tides in December
Campiello but only F. vesiculosus in Artedo). The 1996–January 1997.
lower tidal levels are dominated by Bifurcaria bifur- At each locality and date, 2 sites of the 8 previously
cata, Chondrus crispus and Himanthalia elongata (the selected were randomly chosen and 3 replicates of
latter only in Artedo). each grazing treatment were randomly allocated to
Experimental design. Three grazing treatments each of the 9 quadrats previously selected at each
were considered. Limpets were removed and excluded site. The experiment involved a total of 144 quadrats
from 50 × 50 cm quadrats, hereafter named exclusion (2 localities × 2 seasons × 2 start dates × 2 sites × 3 graz-
quadrats (E), using fences 5 cm high made of plastic ing treatments × 3 replicates). With this design, exper-
coated iron mesh (with a mesh size of 1 cm) attached to imental factors include grazing treatment (G, fixed)
the rock with stainless steel screws and plastic plugs. with 3 levels (E, PC and C), locality (L, random) with
Where the fences did not exactly fit the substratum, 2 levels (Campiello and Artedo), season (Se, fixed)
strips of artificial grass were introduced between the with 2 levels (summer and winter), start date (D, ran-
rock and the mesh to prevent small limpets passing dom), nested in the interaction L × Se and with 2 start
under the fences. Artefacts might occur because of dates per locality and season, and site (Si, random),
alterations of water flow within the exclusion quadrats nested in date and with 2 sites per date at each locality
or increased shading and wetness of the rock surface. and season.
Procedural controls (PC) to test for potential artefacts Two hypotheses were derived from the general mo-
due to barriers consisted of partial fences. Quadrats del that limpets influence the structure of assemblages:
were fenced at the corners, leaving an open space of (1) algal cover increases in exclusion quadrats when
25 cm in the middle of each side. At several randomly compared with both types of controls (C = PC < E) and
chosen corners, strips of artificial grass were in- (2) barnacle cover decreases in exclusion quadrats
corporated. Limpets could move freely in and out of when compared with controls (C = PC > E). The unde-
the quadrats. The environmental conditions in the sirable effect of fences is analysed by testing the
quadrats were more similar to those within the exclu- hypothesis that controls differ (C ≠ PC), both for algae
sion quadrats than to those on natural surfaces (but see and barnacle cover. Temporal variability in the effect
comments and criticism on partial barriers as proce- of limpet removal may be detected after significant
dural controls by Johnson [1992] and Benedetti-Cecchi tests for the interactions of grazing with season (G ×
& Cinelli [1997]). Finally, control quadrats (C) were Se) and grazing with date [G × D(Se × L)]. The spatial
marked with 2 screws on opposite corners and left heterogeneity may be detected after significant tests
otherwise untouched. for the interactions of grazing with locality (G × L) and
To cope with spatial heterogeneity, the experiment grazing with site {G × Si[D(Se × L)]}. Because the re-
was replicated in 2 localities (Campiello and Artedo). maining terms in the model were not directly related to
Within each locality, the grazing treatments were the hypotheses, they were ignored in the ‘Discussion’.
120 Mar Ecol Prog Ser 277: 117–133, 2004
Height above LAT, orientation and slope were re- in cover in individual quadrats [100 × (initial cover –
corded for each quadrat. In addition, an estimation of final cover)/initial cover]. Apart from analyses on the
roughness of the quadrats was obtained by loosely lay- effect of limpet removal on the global assemblage of
ing a thin chain with 4 mm links on the 2 diagonals. algae and barnacles, additional analyses tested the
The excess in length of the sum of the 2 diagonals esti- success of fences in excluding limpets and variations in
mated with the chain in relation to the sum of the true the abundance of trochids.
diagonals of a 50 × 50 cm quadrat was considered an With this design, tests for grazing as a main effect
estimate of the roughness of the rock. and some tests for interaction had a reduced power
Sampling. The experiment ran from June 1996 to (df = 2, 2). Different experimental designs might pro-
January 1998. Proportions of Patella vulgata and P. vide more powerful tests for these factors, but the work
depressa were estimated at the start of manipulations reported here was part of a larger experiment and
from animals removed from a variable number of therefore constrained by the need for a common de-
exclusion quadrats in each locality. Quadrats were sign (Chelazzi et al. 1998a). We pooled non-significant
sampled during low spring tides with the point interactions (p > 0.25) involving random factors to
method using a grid made of double thread to avoid increase the power of the tests involving the grazing
parallax errors and with 49 regularly spaced points. treatment. The sequential decision pooling procedure
Samples were taken at the time of the manipulations, described in Winer (1971) was used. Preliminary tests
15 d later, once a month during the first 3 months and were performed to check if the interaction Grazing ×
then every 3 months until January 1998. To follow Date could be eliminated from the model and pooled
up colonisation by Fucus spp., its cover in exclusion with Grazing × Site. Then, we tested the significance of
quadrats was sampled on 2 additional dates, March Grazing × Locality × Season and Grazing × Locality
and October 1998. Primary and secondary cover of all against the pooled error term. If not significant, the
organisms present in the quadrats was recorded and tests for Grazing × Season and Grazing against the
transformed to percentages. Present but not recorded new pooled error term would have considerably
organisms were assigned a cover of 1%. When pos- increased the power (df = 2, 28). Cochran’s test was
sible, organisms were identified to species in the field, used to test for heterogeneity of variances. When
though some taxonomically difficult groups were required, data were transformed to meet the assump-
always assigned to higher categories (e.g. Cerami- tion of homogeneity of variance.
ales). The number of trochids and limpets, with no The relationship between the proportion of Fucus
distinction of species, was also recorded. Limpets cover and height on the shore as well as the effects of
entering the exclusion quadrats were counted and locality (random factor) and season (fixed) were fitted
removed. by a logistic regression using the SAS Macro program
Data analysis. Hypotheses were tested using analy- GLIMMIX, which iteratively uses the SAS MIXED
sis of variance (ANOVA). The existence of artefacts procedure (SAS 1996). The MIXED procedure imple-
was tested using a priori comparisons (Underwood ments a generalisation of the standard linear model
1997). After significant tests, effects involving grazing, that allows for proper incorporation of random effects.
either as a main effect or in interaction, were split into Model parameter estimates were fitted using the
2 comparisons: Among Controls (i.e. PC vs C) and restricted maximum likelihood method (Litell et al.
Exclusion vs Controls. The former tested for experi- 1996). For further details on the MIXED procedure, see
mental artefacts, while the latter tested for effect of Litell et al. (1996) and SAS (1996).
limpet removal. In some cases, artefacts might exist There is a problem of confounding effects associated
(see ‘Results’) and therefore to estimate the effect of with comparisons of means of treatments started in
limpets we compared Exclusion and Procedural Con- different seasons. For instance, in January 1998, differ-
trol quadrats. Comparisons were not non-independent ences between quadrats established in summer and
and non-orthogonal and therefore the probability of those established in winter could be due to different
Type I error was increased. However, we kept α = 0.05 start seasons as much as to the different time during
for each comparison to avoid excessive Type II errors which the changes occurred (an average of 18 mo for
(Underwood 1997). Algal species were pooled in 2 summer quadrats and 12 mo for winter quadrats). On
groups: (1) Fucus spp. distinguishing between primary the other hand, if means are compared after a fixed
and secondary cover, and (2) all other macroalgae time from establishment of experimental quadrats (e.g.
pooled (only primary cover). For cover of algae, the 12 mo), then differences could be due to the fact that
analysed data were percentage of cover in individual quadrats were sampled at different times of the year
quadrats. For barnacles, because cover at any date is (summer for summer quadrats and winter for winter
dependent on cover at the beginning of the experi- quadrats). If in the first case significant seasonal effects
ment, the analysed data were proportions of change might be due to the time elapsed since the experi-
Arrontes et al.: Effects of grazing by limpets 121
mental manipulations, in the second, seasonal effects 0.05). Slope and roughness of experimental quadrats
could be due to differences in the presence or abun- did not differ among any combination of treatments (no
dance of species at different times of the year. As this significant effects in ANOVA, data not shown). While
problem is unavoidable with the present design, data orientations were not statistically analysed, no obvious
were analysed 12 mo after the establishment of the trend was observed for quadrats under any treatment.
quadrats and at the end of the experiment, and a In Campiello, relative abundances of limpets at the
cautious interpretation of the results was made. In start of the experiment were 81.2 and 18.8% (N = 1267)
other cases, analyses were performed with data at the for Patella depressa and P. vulgata respectively. The
time in which maximum abundance was recorded. mean size (SE) was 1.46 ± 0.012 cm (N = 1029) for
Temporal trends were deduced from visual inspection P. depressa and 1.53 ± 0.035 cm (N = 238) for P. vul-
of graphs. gata. In Artedo, relative abundances of limpets were
90.3 and 9.7% (N = 859) for P. depressa and P. vulgata
respectively, and the mean sizes were 1.51 ± 0.018 cm
RESULTS (N = 776) and 1.47 ± 0.076 cm (N = 83) respectively.
Data on the density of limpets in some quadrats before
Initial conditions manipulations were lost and thus, the initial density of
limpets was estimated from untouched controls at the
Initial cover of barnacles was very variable and var- time of the first sampling. Initial densities were 59.13 ±
ied between 41 and 96%. Significant differences 4.42 (mean ± SE, N = 24) for Campiello and 42.5 ± 3.59
existed among sites (Table 1A). The triple interaction for Artedo. ANOVA revealed that differences in mean
(Grazing × Locality × Season) was also significant. The density were significant only between sites within
height on the shore of experimental quadrats oscillated each date, locality and season (Table 1B).
between 1.17 and 2.64 m above LAT. Despite this wide
range of variation, only significant differences in tidal
height were detected among sites (Table 1A). ANOVA Effects of limpet removal
revealed no significant main effects or interactions,
with associated p always above 0.3. No correlation Fences failed to completely exclude limpets in
existed between the height on the shore of the experi- exclusion quadrats (Fig. 1A); however, the density of
mental quadrats and barnacle cover (r = 0.063, p > limpets was considerably lower in exclusion quadrats
Table 1. Analysis of differences in initial conditions of experimental quadrats. (A) Barnacle cover and tidal height. For cover of
barnacles, variances were homogeneous after arcsine transformation. No transformation was needed for height on the shore.
(B) Density of limpets in unfenced control quadrats. No transformation needed (ns = non-significant, *p < 0.05, ***p < 0.001).
Because the analyses were performed to identify gross differences in initial conditions, pooling was not necessary
Source of variation df Error term MS F p MS F p
(A) Barnacles and tidal height –––––––––– Barnacles ––––––––––– ––––––– Tidal height ––––––
(a) Grazing = G 2 (d) 0077.18 1.09 ns 0.033 0.62 ns
(b) Locality = L 1 (h) 1142.88 4.68 ns 0.030 0.07 ns
(c) Season = Se 1 (f) 0238.97 0.28 ns 0.021 0.05 ns
(d) G × L 2 (i) 0071.10 3.23 ns 0.054 0.99 ns
(e) G × Se 2 (g) 0010.47 0.09 ns 0.010 0.19 ns
(f) L × Se 1 (h) 0855.84 3.51 ns 0.427 0.96 ns
(g) G × L × Se 2 (i) 0117.87 5.36 * 0.054 0.99 ns
(h) Date = D(L × Se) 4 (j) 0243.99 1.11 ns 0.446 1.46 ns
(i) G × D (L × Se) 8 (k) 0021.98 0.29 ns 0.055 0.98 ns
(j) Site = Si[D(L × Se)] 8 (l) 0219.80 5.01 *** 0.306 4.65 ***
(k) G × Si[D(L × Se)] 160 (l) 0075.85 1.73 ns 0.056 0.85 ns
(l) Residual 960 043.88 0.066
(B) Limpets
(a) Locality = L 1 (d) 3316.69 2.59 ns
(b) Season = Se 1 (c) 0165.02 0.19 ns
(c) L × Se 1 (d) 0858.62 0.67 ns
(d) Date(L × Se) 4 (e) 1281.06 2.03 ns
(e) Site[D(L × Se)] 8 (f) 0630.85 3.01 *
(f) Residual 320 0209.50
122 Mar Ecol Prog Ser 277: 117–133, 2004
than in the controls. Most of the limpets found within cant interaction reflected differences in the pooled
the fences were of small size and their numbers fol- abundance in both types of controls.
lowed seasonal variations in abundance in control At the time of highest density (November 1997),
quadrats (data not shown). The number of limpets in trochids were more abundant in exclusion quadrats
exclusion quadrats exhibited an increased trend than in controls (Fig. 1B). On some sampling dates
towards the end of the experiment, possibly reflect- (summer months, data not shown) virtually no animals
ing damage to the fences and previously unnoticed were found in control quadrats, while they were still
recruitment. At the time of highest abundance of present in exclusion quadrats. Large differences in
limpets (May 1997), differences existed between ex- density also existed between sites. ANOVA was per-
clusion and procedural control quadrats, but also be- formed for data collected in November 1997 (Table 2).
tween both types of control quadrats (Fig. 1A). Aver- Apart from the significant site effects, the analysis
age density was higher in procedural controls than in revealed significant differences between exclusion
untouched control quadrats. Apparently, fences quadrats and controls. No significant differences ex-
attracted limpets. Large differences among sites are isted between the 2 types of controls, though a trend of
also evident in Fig. 1A. ANOVA revealed that the larger density in the procedural control quadrats as
effects of grazing and site were significant (Table 2). compared to untouched controls was observed. In fact,
Analyses performed 3, 6 and 12 mo after manipula- if ANOVA was exclusively performed with control
tions (data not shown) rendered similar results, quadrats (data not shown), a significant difference
though the interaction Grazing × Site was also signif- between the 2 types of controls existed. As in the case
icant. Because the number of limpets in exclusion for limpets, trochids appeared to be attracted by
quadrats was fairly constant across sites, this signifi- fences.
Exclusion
A. Limpets
Procedural Control
Control
80 125
Campiello Artedo
100
60
75
40
50
No. of animals per quadrat
20
25
0 0
Site 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2
Date D1 D2 D1 D2 D1 D2 D1 D2
Season Summer Winter Summer Winter
B. Trochids
30 60
Campiello Artedo
20 40
10 20
0 0
Site 1 2 1 2 1 2 1 2 1 2 1 2 1 2 1 2
Date D1 D2 D1 D2 D1 D2 D1 D2
Season Summer Winter Summer Winter
Fig. 1. Mean abundance of (A) limpets in the 3 grazing treatments in May 1997 and (B) trochids in November 1997. Left panels
are data from all sites pooled (N = 48). Right panels are abundance at individual sites. In all cases, data are mean ± SE
Arrontes et al.: Effects of grazing by limpets 123
Table 2. Analysis of differences in density of limpets (May 97) and Trochids (Nov 97). Variances were homogeneous after square
root transformation. In both cases, the value for 1 control quadrat lost in Artedo was replaced by the mean of the remaining
2 quadrats of the same treatment and site, and 1 df subtracted from the residual (ns = non-significant, *p < 0.05, ***p < 0.001)
(a) Source of df Error Limpets Trochids
(a) variation term MS F p MS F p
(a) Grazing = G 2 364.01 77.091
Exclusion vs
procedural control 1 Pooled-2 609.34 360.2511 *** 94.151 35.221 ***
Among controls 1 Pooled-2 118.45 5.00 * 3.45 1.29 ns
(b) Locality = L 1 (h) 110.55 0.05 ns 5.13 0.83 ns
(c) Season = Se 1 (f) 110.24 0.01 ns 23.721 50.621 ns
(d) G × L 2 Pooled-1 111.22 0.38 ns 1.48 0.52 ns
(e) G × Se 2 Pooled-2 110.62 0.37 ns 2.23 0.83 ns
(f) L × Se 1 (h) 120.78 1.90 ns 0.47 0.08 ns
(g) G × L × Se 2 Pooled-1 112.38 1.42 ns 1.94 0.68 ns
(h) Date = D(L × Se) 4 (j) 110.95 3.71 ns 6.18 0.69 ns
(i) G × D(L × Se) 8 (k) 111.92 1.24 ns 3.54 1.42 ns
(j) Site = Si[D(L × Se)] 8 (l) 112.95 2.56 * 9.01 3.38 *
(k) G × Si[D(L × Se)] 16 (l) 111.55 1.34 ns 2.48 0.93 ns
(l) Residual 95 111.15 2.66
(l) Pooled-1 error term: (i) + (k) 24 1.67 2.83
(l) Pooled-2 error term: (d) + (g) + (i) + (k) 281 1.69 2.67
Effect of limpet removal on macroalgae concluded that departure from normality had little
effect. Early colonisation, estimated as increases in
The most abundant algal species in exclusion primary cover, varied between localities and seasons
quadrats were Fucus spp. One year after manipulation, (significant interaction in ANOVA for primary cover,
several quadrats exhibited a secondary cover above Table 3). Colonisation was very rapid in Artedo in sum-
95%. Although all specimens that could be safely iden- mer quadrats and similar in both seasons in Campiello,
tified were F. vesiculosus, many small specimens re- and in Artedo in winter (Fig. 2). No other effect was
mained unidentified and we chose to keep the generic significant. Twelve months after manipulation, ANOVA
denomination. Except for 1 site in Artedo, in which revealed that the only significant effect was site. As
Fucus colonised both types of control quadrats, mea- shown in Fig. 3, sites within the same time period,
surable cover of Fucus only appeared in exclusion locality and season exhibited large differences in aver-
quadrats (Fig. 2). Only these quadrats were analysed age cover (e.g. more than 80% and less than 10% for
for differences in cover. The analysis is simplified by sites on one winter date in Artedo). Although mean
dropping the main effect of grazing and its interactions secondary cover appeared to be higher in summer
from the model and testing for spatial and temporal than in winter quadrats (Fig. 2), no effect of season was
effects on the growth of Fucus in exclusion quadrats. detected. In January 1998, with quadrats being 18 and
There were problems with analyses of cover of Fucus. 12 mo old, the results of ANOVA were identical (data
Even after transformation of data, heterogeneity of not shown). In October 1998, angular transformation
variances and gross departures from normality existed rendered homogeneous variances and almost no bi-
for some of the sampling dates. Lack of normality was modal distribution. Again, ANOVA revealed that site
a consequence of obvious bimodality of data within the was the only significant effect. This result suggests that
exclusion treatment, with a fraction of the quadrats trends inferred from previous analyses may be close to
lacking measurable cover of Fucus and a fraction with reality. However, differences among sites were smaller
large secondary covers. The analyses were made any- (Fig. 3).
way and therefore their results must be used to suggest Colonisation by Fucus was influenced by the height
possible trends rather than to construct unambiguous on the shore of individual quadrats. More Fucus
conclusions. However, non-parametric analyses of appeared in lower than in higher quadrats (Fig. 4,
variance were performed where clear bimodal data Table 4). No effect of locality existed but summer
existed (DISTLM v.2, Anderson 2001, 2003, McArdle & quadrats were significantly affected by height.
Anderson 2001). F -ratios and associated probabilities Excluding Fucus spp., a total of 27 taxa of algae were
were almost identical to those obtained with para- identified in the experimental quadrats. Their relative
metric ANOVAs (data not shown) and therefore, it is abundance was estimated as the percentage of the
124 Mar Ecol Prog Ser 277: 117–133, 2004
20
Summer. Campiello Winter. Campiello
15
Fucus primary cover (%)
10
5
0
20
Summer. Artedo Winter. Artedo
15
10
5
0
J J A SOND J FMAMJ J A SOND J FMAMJ J A SON J J A SOND J FMAMJ J A SOND J FMAM J J A SON
1996 1997 1998 1996 1997 1998
Exclusion Procedural control Control
100
Summer. Campiello Winter. Campiello
75
Fucus secondary cover (%)
50
25
0
100
Summer. Artedo Winter. Artedo
75 Fig. 2. Fucus spp. Changes in
cover. Top panels: primary cover
50
in exclusion quadrats alone.
25 Bottom panels: secondary cover.
Exclusion quadrats were sam-
0 pled for secondary cover on 2
additional dates, March and
J J A SOND J FMAMJ J A SOND J FMAMJ J A SON J J A SOND J FMAMJ J A SOND J FMAMJ J A SON October 1998. In all cases, data
1996 1997 1998 1996 1997 1998 are mean ± SE (N = 12)
Table 3. Analysis of differences in Fucus cover. Primary cover was analysed 6 mo after manipulation. Variances were still het-
erogeneous (Cochran’s C = 0.36, 0.01 < p < 0.05) after arcsine transformation. For analyses of secondary cover, arcsine trans-
formation rendered homogeneous variances though gross departure from normality occurred 12 mo after manipulations. For
October 1998, variances were homogeneous and close to normality after arcsine transformation (ns = non-significant, *p < 0.05,
**p < 0.01). Pooling of non-significant error terms did not alter the significance of tests for locality, season and their interaction,
and is not presented
Source of df Error Primary Secondary
variation term 6 mo 12 mo Oct 98
MS F p MS F p MS F p
(a) Locality = L 1 (d) 355.32 7.35 ns 1406.53 0.49 ns 1666.74 0.53 ns
(b) Season = Se 1 (c) 123.12 0.23 ns 6203.44 2.93 ns 1152.32 6.25 ns
(c) L × Se 1 (d) 516.89 10.691 * 2119.39 2.54 ns 1124.39 0.02 ns
(d) Date(L × Se) 4 (e) 148.38 0.21 ns 1835.69 0.40 ns 1247.23 0.80 ns
(e) Site[D(L × Se)] 8 (f) 223.43 2.02 ns 2083.85 3.83 ** 1560.72 2.51 *
(f) Residual 32 110.63 1543.99 1621.53
Arrontes et al.: Effects of grazing by limpets 125
Artedo Campiello
Summer Summer
Winter Winter
100
Fucus (%)
80
60
40
20
0
1.00 1.50 2.00 2.50
Tidal height (m)
Fig. 4. Fucus spp. Relationship between secondary cover of
individual exclusion quadrats 12 mo after the start of the
experiment and height on the shore
ences; possibly due to the large site effect. The pattern
of variation in abundance was similar for summer and
Fig. 3. Fucus spp. Mean secondary cover at individual sites winter quadrats in Campiello (Fig. 6). Highest cover
12 mo after the start of the experiment and in October 1998 was observed 10–11 mo after manipulations. In addi-
(only exclusion quadrats). Data are mean ± SE (N = 3) tion to low cover, no clear pattern was observed in
summer quadrats in Artedo. It could be hypothesised
that rapid growth of Fucus in these quadrats prevented
number of interceptions of individual taxa in relation to growth of other algae.
the total number of interceptions of all algae (Fucus
excluded), in all quadrats, on all sampling dates. Most
of the identified taxa were recorded sporadically. The Effect of limpet removal on barnacles
soft encrusting species Ralfsia verrucosa and the blue-
green Rivularia bullata were the most abundant taxa There was an overall trend of barnacles to decrease
and altogether comprised more than 75% of all obser- with time (Fig. 7). The decrease, however, was more
vations. Groups with more than 1% of relative abun- marked in exclusion quadrats. Apparent differences
dance of algae are presented in Fig. 5. also existed between seasons (more barnacles being
Measurable algal growth almost exclusively occur- lost in exclusion summer quadrats) and between sites
red in exclusion quadrats (Fig. 6) and therefore, only within each date, locality and season. Data were
these were analysed for differences in algal cover. analysed 12 mo after manipulations and in January
On the dates of highest abundance (May
and October 1997 for summer and winter
quadrats in Campiello, and November Table 4. Analysis of the relationship between the proportion of Fucus cover
1997 for both seasons in Artedo), the aver- and height on the shore. Logistic regression using SAS GLIMMIX (SAS 1996)
age percentage of cover varied between
more than 40% in some sites to less than Test of effects df p Parameter
Likelihood ratio test (χ2) estimates (SE)
5% in others (Fig. 6). The analysis was per-
formed with data from these sampling Location (random) 0.60 1 > 0.750<
dates. ANOVA revealed that only the site
was significant (Table 5). Although im- F-value Summer:
portant differences appeared to exist be- Season (fixed) 7.71 1, 44 0.008 1.421 (1.813)
Tidal height (fixed) 5.52 1, 44 0.023 –2.449 (0.512)
tween summer quadrats in Campiello and Intercept 3.458 (1.813)
Artedo, ANOVA failed to detect differ-
126 Mar Ecol Prog Ser 277: 117–133, 2004
sa 1998. For this date, an additional analysis was per-
rru a
ve alfsi
co
60 formed with summer quadrats. After 12 mo, variances
R
were still heterogeneous after angular transformation
Relative abundance (%)
and the comment made for Fucus cover is applicable.
45
p.
Analyses for both 12 mo and January 1998 data re-
sp
vealed a significant main effect of grazing, with exclu-
ix
hr
s
a
es
ium inc sum
inn p.
an
i fi d
lot
sion quadrats losing more barnacles than procedural
lm tus
id
30
a p sp
sp rust
Ca
at
ho
um rtuo
he ella
a
lum
controls and no differences among controls (Table 6A).
p.
int
un rph
)
xa
ph m to
on s st
sp
p.
sp usil
o
ta
This main effect, however, is not interpretable due to
ho ella
m
de
m rp u
u
p
5
op
15
yll
Lit hyll
(1
p.
Ul ium
the significant interactions. The site and the more rele-
uin
Ne oca
er
rs
m
ali
m
p
do
t
he
ho
lid
En
Os
t
ra
va vant interaction of grazing with the site were also
as
Au
Ge
Ot
Ce
Lit
0 M significant. In some sites, but not in others, the removal
Fig. 5. Relative abundance of macroalgae (Fucus excluded) in
of limpets led to a larger reduction of barnacles in
all quadrats at all sampling dates (see text). Only taxa with a relation to procedural controls. Both analyses also re-
relative abundance above 1% are identified vealed that differences among controls existed. These
differences were not consistent across
sites and though there was a trend in
some sites for procedural controls to
lose more barnacles than untouched
controls, in others, the opposite or no
trend was observed. As for the density
of limpets, the analyses suggested that
some artefacts caused by fences might
have occurred. Significant effects of
locality existed for the 12 mo data, with
more barnacles being lost in Campiello
than in Artedo. For the January 1998
data, the interaction between grazing
and season was significant, more
barnacles being lost in exclusion than
in controls treatments in quadrats
established in summer, with a smaller
difference in winter quadrats (Fig. 7).
No significant interaction between
differences among controls with sea-
sons was detected. Loss of barnacles
was also evident if only summer
quadrats in January 1998 were consid-
ered (Fig. 7). ANOVA for these data
rendered similar significant results
(Table 6B), but no significant differ-
ences between controls were found.
Loss of barnacles due to the exclu-
sion treatment was correlated with
cover of Fucus of individual quadrats
(Fig. 8). The proportion of barnacle
cover lost 12 mo after manipulations
was positively related to cover by
Fucus (r = 0.523, df = 46, p < 0.01). Note
that an inverse relationship existed
Fig. 6. Top panel: changes in the primary cover of macroalgae (Fucus between height on the shore and the
excluded) during the experiment (N = 12). Arrows indicate dates at which percentage of cover of Fucus. Simi-
maximum cover was observed at each locality and season. These dates were
used for ANOVA. Bottom panel: mean abundance of macroalgae at individual
larly, loss of barnacles was also in-
sites on the sampling dates with maximum cover (N = 3). In all cases, data are versely related to height on the shore
mean ± SE of the quadrats (data not shown, r =
Arrontes et al.: Effects of grazing by limpets 127
Table 5. Analysis of differences in cover by macroalgae quadrats under each grazing treatment in which
(Fucus excluded). Only exclusion quadrats were analysed. whelks were observed at least once is compared,
Variances were homogeneous and data were not transformed
whelks occurred more in exclusion than in control
(ns = non-significant, ***p < 0.001). Pooling of non-significant
error terms did not alter the significance of tests for locality, quadrats (Table 7). No significant differences were
season and their interaction, and is not presented observed between controls. Data suggested a rela-
tionship between secondary cover of Fucus 18 mo
Source of df Error MS F p after manipulation and the accumulated number of
variation term whelks counted in individual exclusion quadrats
manipulated in summer (Fig. 9B). While at high
(a) Locality = L 1 (d) 1096.81 4.08 ns covers of Fucus whelks were either fairly common or
(b) Season = Se 1 (c) 1144.44 0.10 ns rare, at low covers whelks were always rare. There
(c) L × Se 1 (d) 1455.70 1.69 ns could also be a relationship between the accumulated
(d) Date(L × Se) 4 (e) 1269.10 0.24 ns
number of whelks in exclusion quadrats and the
(e) Site[D(L × Se)] 8 (f) 1121.82 12.701 ***
proportion of barnacle cover lost (Fig. 9A). Again,
(f) Residual 321 1188.33
while both a small or large proportion of cover could
–0.370, df = 46, p < 0.05). For summer
quadrats in January 1998 (18 mo old) the
relationships still held (Fucus, r = 0.712,
df = 22, p < 0.01; height, r = –0.418, df =
22, p < 0.05).
The loss of barnacle cover may also
be a consequence of another indirect
effect. Specimens of the whelk Thais
lapillus were occasionally found within
the experimental quadrats. Occurrence
was very irregular, whelks being ob-
served in a low proportion of the visits
to individual quadrats and, when pre-
sent, in variable numbers, ranging from
1 to 37 specimens in 1 procedural con-
trol (Table 7). Although this observation
precluded any formal statistical com-
parison of the effect of grazing mani-
pulation as undertaken above, some
exploratory description is possible. If
all animals counted in all quadrats at
all sampling dates under each grazing
treatment are pooled, a trend can be
observed (Fig. 9A, Table 7). More ani-
mals appeared in exclusion than in
control quadrats. In addition, if data
are pooled for 3 periods, 0 to 6, 7 to 12
and 13 to 18 mo after manipulations,
then it can be observed that differences
became relevant in late sampling dates
(Fig. 9A). If the number of individual
Fig. 7. Top panels: changes in barnacle cover
(N = 12). Bottom panels: losses in barnacle
cover at individual sites 12 and 18 mo after the
start of the experiment (N = 3). In all
cases, data are mean ± SE
128 Mar Ecol Prog Ser 277: 117–133, 2004
be lost at low numbers of whelks, when whelks were Table 7. Total number of whelks Thais lapillus recorded in
fairly common, the loss of barnacles was always experimental quadrats from June 1996 to January 1998, range
of abundance in individual quadrats and number of quadrats
above 50%.
in which whelks were recorded at least once (occurrence).
Chi-square tests for 2 × 2 contingency tables (df = 1) are for
differences in occurrence of whelks under different combina-
DISCUSSION tions of treatments. There were 48 exclusion quadrats, 48 pro-
cedural controls and 47 controls (1 control quadrat was lost)
(ns = non significant, ***p < 0.001)
The experimental results supported the hypotheses
tested. Changes in the assemblages of the exclusion
quadrats were greater than those in both types of con- Treatment Total number Range Occurrence
trols. Exclusion promoted macroalgal growth and
Exclusion 406 1–26 32
induced the loss of barnacle cover. The ability of lim-
Procedural control 80 1–37 14
pets to control algal growth and to influence the abun- Control 20 1–5 8
dance of other animal species is not at all surprising as
χ2 tests
there is overwhelming evidence for this worldwide
Exclusion vs controls 25.68***
(Underwood & Jernakoff 1981, see review by Hawkins Among controls 1.97 ns
& Hartnoll 1983). The relevant aspects are the scales
Table 6. Analyses of differences in the percentage of cover lost by barnacles. After arcsine transformation, variances were still
heterogeneous for 12 mo data. Variances were homogeneous after transformation for data in January 1998. One missing replicate
and 1 erroneous value were replaced by the mean of the remaining 2 replicates of the same treatment at the same sites and 2 df
subtracted from the residual (ns = non-significant, *p < 0.05, **p < 0.01, ***p < 0.001)
Source of variation df Error term MS F p MS F p
(A) All experimental quadrats ––––––––– 12 mo –––––––– –––––––– Jan 98 –––––––
(a) Grazing = G 2 3262.73 6436.31
Exclusion vs procedural control 1 Pooled-2 3269.84 8.28 ** 5923.25 10.681 **
Among controls 1 Pooled-2 1433.58 1.10 ns 1136.57 2.05 ns
(b) Locality = L 1 (h) 15107.961 16.141 * 4995.27 6.14 ns
(c) Season = Se 1 (f) 5096.00 8.56 ns 1470.70 0.61 ns
(d) G × L 2 Pooled-1 1459.49 1.12 ns 1166.98 0.27 ns
(e) G × Se 2 Pooled-2 1597.06 1.51 ns 2041.09
Exclusion vs procedural control × Se 1 Pooled-2 2816.24 5.08 *
Among controls × Se 1 Pooled-2 1118.31 0.03 ns
(f) L × Se 1 (h) 1595.61 0.64 ns 1774.63 0.95 ns
(g) G × L × Se 2 Pooled-1 1165.80 0.41 ns 1178.62 0.13 ns
(h) Date = D(L × Se) 4 (j) 1936.26 1.97 ns 1813.64 1.34 ns
(i) G × D(L × Se) 8 (k) 1139.61 0.26 ns 1319.31 0.41 ns
(j) Site = Si[D(L × Se)] 8 (l) 1474.55 2.12 * 1608.16 3.05 **
(k) G × Si[D(L × Se)] 161 1543.29 1780.31
Exclusion vs procedural control × [D(L × Se)] 8 (l) 1586.30 2.62 * 1107.75 5.56 ***
Among controls × Si[D(L × Se)] 8 (l) 1803.17 3.59 ** 1715.40 3.59 **
(l) Residual 941 1223.55 1199.20
Pooled-1 error term: (i) + (k) 241 1408.76 1626.64
Pooled-2 error term: (d) + (g) + (i) + (k) 281 1395.03 1554.66
(B) Summer quadrats in January 1998 (18 mo)
(a) Grazing = G 2 7763.861
Exclusion vs procedural control 1 Pooled-2 8454.021 12.211 **
Among controls 1 Pooled-2 721.70 1.04 ns
(b) Locality = L 1 (d) 917.85 1.71 ns
(c) G × L 2 Pooled-1 189.02 0.24 ns
(d) Date = D(L) 2 (f) 538.15 0.65 ns
(e) G × D(L) 4 (g) 491.35 0.54 ns
(f) Site = Si[D(L)] 4 (h) 822.24 3.15 *
(g) G × Si[D(L)] 8 918.36
Exclusion vs procedural control × Si[D(L)] 4 (h) 1406.931 5.38 **
Among controls × Si[D(L)] 4 (h) 485.49 1.86 ns
(h) Residual 46 261.31
Pooled-1 error term: (e) + (g) 12 776.02
Pooled-2 error term: (c) + (e) + (g) 14 692.17
Arrontes et al.: Effects of grazing by limpets 129
of variability of this effect, the nature of the A
interactions behind the measurable effects and Exclusion
200
Procedural control
Total no. of whelks
how this variability and interactions change
over large spatial scales. Control
150
An identical experimental design was used to
evaluate the effects of grazing on the structure 100
of moderately exposed rocky shores on the west
coast of Sweden (Lindegarth et al. 2001). How- 50
ever, a formal comparison of the results from
0
both experiments, although possible, is not 0-6 7-12 13-18
useful. Differences in the composition of the Period (months)
intertidal assemblages and contrasting environ-
mental conditions preclude any sensible inter- B 18 months Campiello Artedo
pretation of a common statistical analysis since
Proporion of barnacle
60 1.0
any observed difference might be a conse-
No. of whelks
quence of multiple factors. The shores of the 0.8
cover lost
north of Spain and the west of Sweden differ in 40
0.6
the main key grazer: limpets in Spain and Litto-
0.4
rina littorea in Sweden (Cervin & Åberg 1997, 20
0.2
Viejo et al. 1999). Dominant algal groups in
Sweden include filamentous and crustose red 0 0.0
algae and ephemeral greens (Lindegarth et al. 0 20 40 60 80 100 0 20 40 60
2001), while virtually no macroalgae are pre- Fucus (%) No. of whelks
sent at mid-tidal levels in northern Spain
Fig. 9. (A) Total number of whelks counted in different periods during
(Anadón 1983). In addition, obvious different the experiment. For the initial period (0 to 6 mo) data are accumulated
thermal and radiation regimes exist in both numbers from 8 sampling dates. For the periods 7 to 12 and 13 to
places, with ice-scouring being a fairly common 18 mo, data are accumulated from 3 sampling dates. The period 13
event in Sweden. Tides in northern Spain are to 18 mo is for quadrats manipulated in summer. (B) Relationship
between the accumulated number of whelks and the percentage of
cover of Fucus (left) and between the proportion of barnacle loss and
Campiello Artedo accumulated number of whelks (right). In both cases, data are from
summer individual quadrats after 18 mo of experimental manipulation
Summer Summer
Winter Winter
12 months semi-diurnal with a maximum range of 4.3 m, while in
0.8 Sweden the tidal range is very narrow and often
0.6 dwarfed by unpredictable sea-level variations (Johan-
Proportion of barnacle cover lost
0.4 nesson 1989). However, qualitative comparisons are
0.2 possible. Site effects were important in both studies.
0.0
The role of grazing in the change of the structure of
assemblages across dates and sites was weak and
-0.2
inconsistent in Sweden, and it is concluded by Linde-
garth et al. (2001) that grazing is less important there
18 months than in semi-exposed shores in other parts of the
1.0
world. Our results support this conclusion as we show
0.8 that grazing affected the structure of assemblages by
0.6 limiting the development of ephemerals and algal
0.4
canopies.
Other studies of the grazing activity of limpets over-
0.2
lap with the experiment reported in this paper. Using
0.0 wax discs placed on the rock surface in Campiello and
0 20 40 60 80 100 Artedo, Jenkins et al. (2001) estimated that an average
Fucus (%) of 26% of rock surface was scraped by limpets at
Fig. 8. Relationship between the proportion of barnacle cover
natural densities in 2 wk, without any seasonal trend
lost and cover of Fucus in individual exclusion quadrats 12 and with similar values for both localities. This amount
and 18 mo after the start of the experiment may appear low and, in fact, it is significantly lower
130 Mar Ecol Prog Ser 277: 117–133, 2004
than estimated rates for southern England or southern exclusion quadrats reverted to a situation similar to
Portugal (Jenkins et al. 2001). However, if we consider that before the manipulations. The persistence of
that an escape size for Fucus is only attained several dense cover of macroalgae depends upon the con-
months after the settlement of zygotes, then it may be tinuity of the manipulation and thus the grazer- and
enough to prevent the development of a Fucus canopy. barnacle-dominated assemblage and the algal patches
In addition, grazing by limpets is not a random process cannot be considered alternate stages at these tidal
(e.g. Chelazzi et al. 1998b) and the spatial organisation levels (Petraitis & Dudgeon 1999).
of foraging might be related to the availability of food. The second prediction, decreased barnacle cover
The estimated 26% of rock surface grazed in Jenkins after limpet removal, was also supported by the exper-
et al. (2001) is a mean value but individual wax discs imental results. However, the effect was measurable
presented large variability ranging from 0 to virtually only after 12 mo from the beginning of the manipula-
100% of the surface scraped. If the density of limpets tions. The lag in the effect is interpreted as only when
were well below natural levels, the grazing pressure algal cover was important and persisted for some
might not be enough to prevent algal growth, in which months did barnacles start to decline. Another relevant
case the probability of escape of individual algal pro- result is that the effect of the grazing treatment is site-
pagules would increase. No estimates on the efficiency dependent and presumably reflects site effects in the
of scraping in removing algal propagules was avail- growth of Fucus and other macroalgae. Results were
able, but reports on the effects of limpets on the consistent across seasons, dates and localities. The
microflora growing on the rocks in Australia (Under- most recurrent result was that small-scale hetero-
wood & Jernakoff 1981) or studies on the anatomical geneity was important (significant site effects). Site
structure of the radula (Steneck & Watling 1982) effects might reflect environmental differences among
suggest that scraping may be very efficient. Radula sites. Before manipulations, sites exhibited differences
strokes virtually collect all algae from the rock surface. in barnacle cover and abundance of limpets. In addi-
The experimental design allowed us to test the con- tion, they had different mean tidal heights but were
sistency of the effect of grazing spatially and over time. similar in orientation, slope and roughness. Other fac-
No effect of start season or date was detected, as the tors, which could potentially influence the composition
interaction of grazing with these 2 factors was usually and dynamics of the assemblage, such as the prevail-
not significant. This is possibly due to the low variabil- ing direction of waves or exposure were not investi-
ity in pressure grazing by limpets over time (Jenkins gated. Site effects may be generated by multiple fac-
et al. 2001). However, the season of removal had a tors acting independently or synergistically. Tidal
significant effect on macroalgal growth at 1 locality, height alone may influence the growth rate of algal
resulting in a more rapid colonisation in Artedo during propagules (Raffaelli & Hawkins 1996). The direction
summer than in winter. This may be related to the of incoming waves and the distance to sources of
availability of propagules after the removal of limpets, propagules may affect not only the rate of develop-
in a similar mode as season for scraping influenced ment of algal canopies but also the possibility that
colonisation of experimental patches by Fucus in some algae (e.g. Fucus) could even reach individual
British Columbia (Kim & DeWreede 1996). Twelve and experimental quadrats (Arrontes 2002).
18 mo after manipulation, however, the significant There is another effect related to limpet removal: in-
interaction vanished. In part, this could be due to the creased densities of trochids (mainly Gibbula umbili-
method of estimation of macroalgal abundance. By calis). In this case, a clear interpretation is not possible.
estimating the percentage of cover of macroalgae, The increased density might be an indirect effect
differences in abundance under high and low recruit- caused either by the increase of food on the rock sur-
ment are only detected at early stages. After some face or by the growth of macroalgae, but no relation-
time, as individual plants grow, secondary cover may ship between abundance of algae and trochids was
be similar even with very different density of Fucus found. Alternatively, this increase in trochid density
plants. However, site effects were significant and might be an artefact. Grazers have been observed to be
surely reflect differences among sites (such as tidal attracted by fences and open cages in previous studies
height) independently of experimental manipulations. (e.g. Underwood 1980). A similar trend of increased
Apart from Fucus, other algae were abundant in some density in procedural controls was also observed in
quadrats, mainly Ralfsia-like soft encrusting species limpets and whelks. Some evidence of artefacts also
and ephemerals. In general, their abundance de- appeared in barnacle cover. Potential artefacts were
creased as the Fucus canopy developed. not always statistically significant, but in most cases
After the last sampling date in October 1998, all differences between both types of controls were always
fences were removed. Limpets quickly reinvaded the in the same direction. Significant effects of procedural
experimental quadrats and after 3 yr, all identifiable controls, even with clear exclusion effects, may gener-
Arrontes et al.: Effects of grazing by limpets 131
ate results that are not logically interpretable (Under- nipulation of abundance (e.g. presence/absence) of
wood 1997). In general, however, and due to the mag- the species assumed to exert the indirect effect and
nitude of the change in exclusion quadrats vs proce- (2) manipulation of the abundance of the third species
dural controls, we do not think that artefacts were directly responsible for changes in the abundance of
strong enough to invalidate the main conclusions. the focal species. Additional treatments should control
Both direct and indirect effects are responsible for for artefacts associated with manipulations. To con-
the observed results. The increase of algae is surely an clude that a decrease in barnacle cover is not a direct
immediate consequence of the cessation of limpet consequence of limpet removal but of the growth
grazing and is interpreted as a direct effect. Several of macroalgae, 2 additional experimental treatments
lines of evidence, however, suggest indirect effects on should consider the growth of macroalgae while
barnacles: (1) Limpets have been shown to remove limpets were still present, and the removal of limpets
juvenile and larvae of barnacles, either by accidental with no growth of algae. No experiments of this kind
or deliberate ingestion or by dislodging specimens were performed in this work to identify indirect effects
while crawling over the barnacles (e.g. Wootton 1993). and it is difficult to see how these treatments could be
If a measurable direct effect on barnacles existed it set up. However, considering the reported evidence
should be negative and thus removal of limpets should based on correlations and life-cycles, we are confident
promote an increase in barnacle cover. (2) Towards the that indirect effects exist.
end of the experiment, the rate of loss of barnacles was Our results are summarised in Fig. 10. We present
positively related to algal cover in exclusion quadrats. averaged species interactions but in some sites or
The direct causes might include overgrowth of barna- quadrats the picture may be completely different, as in
cles by algae, reduced water flow limiting availability some quadrats no growth of macroalgae occurred and
of food and inhibition of recruitment. (3) Sporadically, barnacle cover remained unchanged. Of course, there
juvenile whelks were found among algae in exclusion are more links in the assemblage than those in Fig. 10.
quadrats more often than in controls. The latter effect For instance, limpets have been shown to have a small
was also observed in previous studies such as the mon- deleterious effect on barnacles, and whelks are often
itoring program of the effects of the Torrey Canyon oil observed to feed on juvenile limpets (B. Martínez, pers.
spill (summarised in Raffaelli & Hawkins 1996). Juve- obs.). In addition, the interactions between canopy Fu-
nile whelks in Campiello and Artedo tend to aggregate cus and limpets may be complex and generate cycles
among algae, crevices or underneath boulders, partic- (Hawkins & Hartnoll 1983, Burrows & Hawkins 1998).
ularly on clear sunny days (B. Martínez, unpubl.). In The graph is also incomplete because other groups are
short, the indirect effect of limpets on barnacles should ignored; the complete assemblage should include
be positive and mediated by the inhibition of algal small littorinids, acari, insects, grapsid crabs, and inter-
growth. Similarly, the effects of algae on barnacles spersed Lychina pygmaea and Mytilus patches with
may be complex and include additional indirect effects their associated fauna. Fig. 10 only represents a subset
mediated by increasing abundance of juvenile whelks. of a more complex assemblage and a conspicuous
Indirect effects appear to be common in intertidal spatial heterogeneity. Finally, the graph would only be
systems. Menge (1995) estimated that 40 to 50% of the complete if the pattern of arrival of algal propagules
changes occurring after perturbations might be due to and larvae could be incorporated. It is evident that
indirect effects. Indirect effects include a great variety
of species interactions but all of them can be grouped
( -)
into 9 types (Menge 1995). Habitat facilitation is one of Limpets
these types and occurs when one species improves the Trochids
(- )
habitat of another species by altering the abundance of ( -)
a third species (Fairweather 1990). Habitat facilitation (+)
(+)
appears to be the type of indirect effect of limpets on Algae
barnacles described in this work. There are many
examples of indirect effects in intertidal communities ( -) ( -) Whelks
(see Menge 1995 for references). For example, in the (- )
San Juan Islands (Washington), removal of macroalgae Barnacles
by chitons enhances the growth of microalgae, which
are the food for acmaeid limpets (Dethier & Duggins Fig. 10. Summary of species interactions at mid-tidal levels in
1984). northern Spain suggested by various evidence in this paper.
Only groups and links investigated in this paper are pre-
Unambiguous identification of indirect effects de-
sented. Continuous lines: direct effects; discontinuous lines:
mands proper experimental designs involving at least indirect effects; grey line: doubtful effect. The thickness of the
2 factors in a crossed design (Anderson 1999): (1) ma- lines relates to intensity of the effect
132 Mar Ecol Prog Ser 277: 117–133, 2004
much additional work is needed for the unambiguous European Marine Science and Technology Conference,
identification and quantification of direct and indirect Lisbon, 23–27 May 1998: Project synopses, Vol 1: Marine-
systems. European Commission DG 12, Luxembourg,
effects, for the identification of the causes of the site
p 361–376
effects and to discover the role of the species ignored Chelazzi G, Santini G, Della Santina P (1998b) Route selection
in this work. in the foraging of Patella vulgata (Mollusca: Gastropoda).
J Mar Biol Assoc UK 78:1223–1232
Dayton PK, Tegner MJ, Parnell PE, Edwards PB (1992) Tem-
Acknowledgements. The experiment was funded by the poral and spatial patterns of disturbance and recovery in a
European Union as a part of the EUROROCK-project, MAS3- kelp forest community. Ecol Monogr 62:421–445
CT95-0012. Dethier MN, Duggins DO (1984) An ‘indirect commensalism’
between marine herbivores and the importance of com-
petitive hierarchies. Am Nat 124:205–219
LITERATURE CITED Dethier MN, Duggins DO (1988) Variation in strong interac-
tions in the intertidal zone along a geographical gradient:
Anadón R (1983) Zonación en la costa asturiana: variación a Washington-Alaska comparison. Mar Ecol Prog Ser 50:
longitudinal de las comunidades de macrófitos en difer- 97–105
entes niveles de marea. Invest Pesq 47:125–141 Fairweather PG (1990) Is predation capable of interacting
Anadón R, Niell FX (1981) Distribución longitudinal de with other community processes on rocky reefs? Aust
macrófitos en la costa asturiana (N de España). Invest Pesq J Ecol 15:453–464
45:143–156 Farrell TM (1991) Models and mechanisms of succession: an
Anderson MJ (1999) Distinguishing direct from indirect example from a rocky intertidal community. Ecol Monogr
effects of grazers in intertidal estuarine assemblages. 61:95–113
J Exp Mar Biol Ecol 234:199–218 Foster MS (1990) Organization and macroalgal assemblages
Anderson MJ (2001) A new method for non-parametric multi- in the Northeast Pacific: the assumption of homogeneity
variate analysis of variance. Aust Ecol 26:32–46 and the illusion of generality. Hydrobiologia 192:21–33
Anderson MJ (2003) DISTLM v.2: a FORTRAN computer pro- Hawkins SJ (1981) The influence of season and barnacles on
gram to calculate a distance-based multivariate analysis the algal colonization of Patella vulgata exclusion areas.
for a linear model. Department of Statistics, University of J Mar Biol Assoc UK 61:1–15
Auckland, New Zealand Hawkins SJ, Hartnoll RG (1983) Grazing of intertidal algae by
Anderson MJ, Underwood AJ (1997) Effects of gastropod marine invertebrates. Oceanogr. Mar Biol Annu Rev 21:
grazers on recruitment and succession of an estuarine 195–282
assemblage: a multivariate and univariate approach. Hawkins SJ, Harnoll RG, Kain JM, Norton TA (1992) Plant-
Oecologia 109:442–453 animal interactions on hard substrata in the north-east
Arrontes J (1993) Nature of the distributional boundary of Atlantic. In: John DM, Hawkins SJ, Price JH (eds) Plant-
Fucus serratus on the north shore of Spain. Mar Ecol Prog animal interactions in the marine benthos. Clarendon
Ser 93:183–193 Press, Oxford, p 1–32
Arrontes J (2002) Mechanisms of range expansion in the Jenkins SR, Arenas F, Arrontes J, Bussell J and 10 others
intertidal brown alga Fucus serratus in northern Spain. (2001) European-scale analysis of seasonal variability in
Mar Biol 141:1059–1067 limpet grazing activity and microalgal abundance. Mar
Benedetti-Cecchi L, Cinelli F (1994) Recovery of patches in an Ecol Prog Ser 211:193–203
assemblage of geniculate coralline algae: variability at Johannesson K (1989) The bare zone of Swedish rocky shores:
different successional stages. Mar Ecol Prog Ser 110:9–18 why is it there? Oikos 54:77–86
Benedetti-Cecchi L, Cinelli F (1997) Confounding in field Johnson LE (1992) Potential and peril of field experimenta-
experiments: direct and indirect effects of artefacts due to tion: the use of copper to manipulate molluscan herbi-
the manipulation of limpets and macroalgae. J Exp Mar vores. J Exp Mar Biol Ecol 160:251–262
Biol Ecol 209:171–184 Kennelly SJ (1987) Physical disturbances in an Australian
Benedetti-Cecchi L, Bulleri F, Cinelli F (2000) The interplay of kelp community. I. Temporal effects. Mar Ecol Prog Ser 40:
physical and biological factors in maintaining mid-shore 145–153
and low-shore assemblages on rocky coasts in the north- Kim JH, DeWreede RE (1996) Effects of size and season of
west Mediterranean. Oecologia 123:406–417 disturbance on algal patch recovery in a rocky intertidal
Boaventura D, Alexander M, Della Santina P, Smith ND, Ré P, community. Mar Ecol Prog Ser 133:217–228
Cancela da Fonseca L, Hawkins SJ (2002) The effects of Lindegarth M, Åberg PA, Cervin G, Nilsson PG (2001) Effects
grazing on the distribution and composition of low-shore al- of grazing on the structure of mid-shore, intertidal as-
gal communities on the central coast of Portugal and on the semblages on moderately exposed rocky shores on the
southern coast of Britain. J Exp Mar Biol Ecol 267:185–206 Swedish west coast. Mar Ecol Prog Ser 212:29–38
Burrows MT, Hawkins SJ (1998) Modelling patch dynamics Litell RC, Milliken GA, Striuo WW, Wolfinger RD (1996) SAS
on rocky shores using deterministic cellular automata. system for mixed models. SAS Institute, Cary, NC
Mar Ecol Prog Ser 167:1–13 McArdle BH, Anderson MJ (2001) Fitting multivariate models
Cervin G, Åberg P (1997) Do littorinids affect the survival of to community data: a comment on distance-based redun-
Ascophyllum nodosum germlings? J Exp Mar Biol Ecol dancy analysis. Ecology 82:290–297
218:35–47 Menge BA (1995) Indirect effects in marine rocky intertidal
Chelazzi G, Hawkins SJ, Åberg P, Arrontes J, Myers A, Paula interaction webs: patterns and importance. Ecol Monogr
J (1998a) Interactions of physical and biological factors 65:21–74
in the surf and swash zone of European rocky shores. Petraitis PS, Dudgeon SR (1999) Experimental evidence for
In: Barthel KG, Barth H, Bohle-Carbonell M, Fragakis C, the origin of alternative communities on rocky intertidal
Lipiatou E, Martin P, Ollier G, Weydert M (eds) Third shores. Oikos 84:239–245
Arrontes et al.: Effects of grazing by limpets 133
Raffaelli D, Hawkins S (1996) Intertidal ecology. Chapman & Underwood AJ (1997) Experiments in ecology. Their logical
Hall, London design and interpretation using analysis of variance. Cam-
Robles CD, Cubit J (1981) Influence of biotic factors in an bridge University Press, Cambridge
upper intertidal community: dipteran larvae grazing on Underwood AJ, Jernakoff P (1981) Effects of interactions
algae. Ecology 62:1536–1547 between algae and grazing gastropods on the structure of
SAS (1996) SAS/SYSTAT software: changes and enhance- a low-shore intertidal algal community. Oecologia (Berlin)
ments through release 6.11. SAS Institute, Cary, NC 48:221–233
Steneck RS, Watling L (1982) Feeding capabilities and limita- Underwood AJ, Petraitis PS (1993) Structure of intertidal
tion of herbivore molluscs: a functional group approach. assemblages in different locations: how can local pro-
Mar Biol 68:299–319 cesses be compared? In: Ricklefs R, Schluter D (eds) Spe-
Stephenson TA, Stephenson A (1972) Life between tidemarks cies diversity in ecological communities. University of
on rocky shores. WH Freeman, San Francisco Chicago Press, Chicago, p 38–51
Underwood AJ (1978) An experimental evaluation of compe- Viejo RM, Åberg P, Cervin G, Lindegarth M (1999) The inter-
tition between three species of intertidal gastropods. active effects of adult canopy, germling density and graz-
Oecologia (Berl) 33:185–202 ing on germling survival of the rockweed Ascophyllum
Underwood AJ (1980) The effects of grazing by gastropods nodosum. Mar Ecol Prog Ser 187:113–120
and physical factors on the upper limits of distribution of Winer BJ (1971) Statistical principles in experimental design,
intertidal macroalgae. Oecologia (Berl) 46:201–213 2nd edn. McGraw-Hill, New York
Underwood AJ (1984) Vertical and seasonal patterns in com- Wootton JT (1993) Indirect effects and habitat use in an inter-
petition for microalgae between intertidal gastropods. tidal community: interaction chains and interaction modi-
Oecologia (Berl) 64:211–222 fications. Am Nat 141:71–89
Editorial responsibility: Antony Underwood (Contributing Submitted: April 29, 2003; Accepted: April 21, 2004
Editor), Sydney, Australia Proofs received from author(s): August 6, 2004