Marine & Freshwater Research

Transcription

1 CSIRO PUBLISHING Marine & Freshwater Research Volume 49, 1998 CSIRO 1998 A journal for the publication of original contributions in physical oceanography, marine chemistry, marine and estuarine biology and limnology All enquiries and manuscripts should be directed to Marine and Freshwater Research CSIRO PUBLISHING PO Box 1139 (150 Oxford St) Collingwood Telephone: Vic Facsimile: Australia ann.grant@publish.csiro.au Published by CSIRO PUBLISHING for CSIRO and the Australian Academy of Science

2 CSIRO 1998 Mar. Freshwater Res., 1998, 49, Estimating spatial variability in developing assemblages of epibiota on subtidal hard substrata T. M. Glasby Centre for Research on Ecological Impacts of Coastal Cities, Marine Ecology Laboratories, A11, University of Sydney, NSW 2006, Australia. Abstract. A nested hierarchical sampling design was used to estimate the scales of natural variability in developing assemblages of subtidal epibiota on rocky reefs. The appropriate spatial scales were needed for sampling to test for environmental impact in this habitat. Sandstone settlement plates were used to mimic the natural substratum. They were designed and deployed in such a way that the effects of any supporting structures were minimized. Differences in recruitment of epibiota were found at all of the spatial scales examined (10s, 100s and 1000s of metres). When differences were found at the smallest spatial scale, they were generally still detected at the two larger scales. The results highlighted the need for adequate small- and large-scale spatial replication for studies of environmental impact. Extra keywords: spatial scale, sessile organisms, fouling, environmental impact assessment, Australia Introduction Artificial surfaces for settlement (also referred to as settlement panels or plates) have been used extensively to study assemblages of epibiota in intertidal and subtidal habitats. Most studies of subtidal assemblages have focussed on succession, competition or disturbance within epifaunal assemblages and examined how these influence the coexistence of fauna and stability of assemblages (Osman 1977, 1978; Sutherland and Karlson 1977 Dean and Hurd 1980; Russ 1980, 1982; Ayling 1981; Kay and Keough 1981; Kay and Butler 1983; Keough 1983, 1984a, 1984b; Butler 1986, 1991). These studies have resulted in numerous descriptions of the patch dynamics of subtidal epifaunal assemblages (e.g. Osman 1977; Connell and Keough 1985; Butler and Chesson 1990). There have been similar studies in algal assemblages (Foster 1975a, 1975b; Kennelly 1983; Kennelly and Underwood 1993). Rarely, however, have settlement plates either been designed to mimic natural substrata (Osman 1977; Todd and Turner 1986) or been constructed from natural substrata (Kennelly 1983; Hixon and Brostoff 1985; McGuinness 1989). Instead, it has been traditional to use easy-to-handle artificial surfaces such as bakelite, Perspex, concrete and wood. Organisms may respond quite differently to these surfaces compared to natural substrata, as has been shown for some intertidal species (McGuinness 1989). This is potentially problematic when artificial surfaces for settlement are used to gain information about recruitment and establishment of epibiota on naturally occurring surfaces. As recognized by Keough (1983, 1984b), many studies of epifaunal assemblages have not specifically addressed processes of recruitment. Clearly, recruitment is important to consider because it has the potential to influence strongly the composition of epibiotic assemblages (Underwood and Denley 1984; Caley et al. 1996). Surprisingly, there is still a paucity of information about spatial and temporal variability in recruitment of subtidal organisms on hard substrata. The few studies of spatial variability in recruitment of subtidal organisms have found large differences in patterns of recruitment over scales of 10s of metres to 10s of kilometres (e.g. Kay and Keough 1981; Keough 1983; Butler 1986). Furthermore, recruitment may vary considerably over a small range of depths (Osman 1977; Hirata 1987) and may be highly variable within and among years (Sutherland and Karlson 1977; Osman 1977, 1978; Keough 1983; Butler 1986). However, since spatial and temporal patterns of larval recruitment are likely to vary considerably among species, use of higher taxonomic levels of classification may be the most meaningful way of describing patterns of recruitment among habitats (Butler 1986). Information about scales of natural variability in recruitment can be crucial for the design of experiments. For example, if recruitment varied naturally over scales of metres but the smallest scale at which sampling was done was 10s of metres, the sampling would be confounded by the natural, smaller-scale variability (Underwood 1991). The study described here was done to identify the appropriate spatial scales at which to sample developing assemblages of subtidal epibiota among different locations as part of a study of environmental impact. Because previous studies have identified spatial differences in the recruitment and development of subtidal epibiota at a number of scales, it was likely that developing assemblages could be very /MF /98/050429

3 430 T. M. Glasby spatially variable between sites in Sydney, Australia. This prediction was tested by comparing the recruitment and development of epibiotic assemblages on new sandstone surfaces at a number of spatial scales. For the purposes of this study, Butler s (1986) definition of a recruit was used. That is, an organism that has settled on the plate and survived to become visible at the time of sampling. Materials and methods Settlement plates Plates (15 mm thick) cut from Hawkesbury sandstone were used as settlement surfaces. All plates had a grain size and texture similar to cleared areas of subtidal rock, the majority of which is also Hawkesbury sandstone. Results of a previous study (Glasby, unpublished) indicated that cm plates would be appropriate. A rigid polyvinyl chloride bracket (10 cm wide) was glued to the back of each plate with an epoxy resin (Megapoxy HT; Vivacity Engineering, Sydney, Australia). This resin is non-toxic to humans and was assumed to be relatively inert. The bracket was bent so that it did not come into contact with the edge of the plate (Fig. 1a). This meant that the front of the plate was essentially free-standing and its surface area was not increased by contact with supporting structures. The base of the bracket had two slots cut out so that the entire unit could be attached vertically to a hardwood beam ( cm) by bolting it onto two stainless steel threaded rods drilled into the side of the beam (Fig. 1b). Each wooden beam was attached to the substratum by screwing two stainless steel self-tapping screws into rawlplugs drilled into the rock. Sampling design Plates were deployed in Pitt Water, a waterway in the southern region of Broken Bay, 30 km north of Sydney, Australia. Pitt Water is bounded by the Ku-ring-gai Chase National Park on the western side and urban areas to the east and south. There are no industrial areas around the waterway and there is no commercial shipping, although it is a popular area for recreational boating. The study sites were adjacent to the National Park and at least 2 km from the nearest human development and were considered therefore to be in a relatively undisturbed area. A fully nested design which incorporated four spatial scales was used. At the largest spatial scale, there were two locations ~ 2 km apart. This represented the distance that control or reference locations may be from a disturbed location in an environmental study. Within each location, there were two sites ~ 300 m apart and within each site were two plots ~ 20 m apart. Three replicate plates were deployed at each plot, one replicate per wooden beam, so that replicates were approximately 3 m apart. Thus, this (a) (b) Fig. 1. (a) side view of sandstone plate with PVC backing; (b) part of a wooden beam showing plates attached vertically by two stainless steel nuts and threaded rod (drilled into side of beam). design enabled the comparison of three spatial scales, 1000s of metres (locations), 100s of metres (sites) and 10s of metres (plots). Seventy-two plates were put out over two days in June (winter) of Three plates were positioned on each beam (each 15 cm apart), one for each time of sampling. Plates were set up in shallow water (~ 0.5 m below Mean Low Water Springs) at each plot. The first set of plates was collected (n = 3 per plot) after 6 weeks. A second sample was collected after 10 weeks (i.e. 4 weeks after the first sample), and the final plates were collected after a total of 19 weeks. These three independent times of sampling were decided on by examining the growth on the plates, thus ensuring that the assemblages were at very different stages of development when sampled. Plates were transported in seawater to the laboratory where they were fixed in a 5% formaldehyde solution neutralized with CaCO 3 and mixed with seawater, then transferred into a solution of 60% ethanol, 10% glycerine and 0.5% formaldehyde in seawater for storage. The plates were examined under a Wild M8 zoom stereo-microscope and only the front surfaces of the plates were sampled. Organisms were identified to a fairly coarse level, in part because of the great taxonomic uncertainty of many subtidal organisms in Australia (Keough and Butler 1995). Percentage covers of sessile organisms were estimated by use of a grid of 100 regularly spaced points which sampled to within 1 cm of the edges of the plates (i.e. a cm area) to avoid edge effects. Abundances of solitary species were not estimated because they had been found to correlate well with percentage cover for each taxon (Glasby, unpublished). Analyses of data Data were analysed by univariate and multivariate techniques. For analyses of variance (ANOVA) all factors were considered random. Variances were tested for homogeneity by Cochran s C-test (P = 0.05), and if data were heterogeneous α was set at 0.01 for the ANOVA (Underwood 1981). Non-significant factors (at P = 0.25) were pooled post hoc to increase the power of other tests in the model (Winer 1971; Underwood 1981). Means were compared by Student Newman Keuls (SNK) tests. Components of variation were calculated for each of the three spatial scales (Underwood 1997). Negative variance components have been presented. These indicate that there is essentially no variation among the levels of the factor in question, but that the true value has been underestimated. Consequently, the variation between levels of higher factors is also underestimated, whereas that between levels of lower factors is overestimated. Non-parametric multivariate techniques used the PRIMER software package (Plymouth Marine Laboratory, UK). Data were double square-root transformed and Bray Curtis similarity matrices (Bray and Curtis 1957) were calculated according to the recommendations of Clarke and Green (1988) and Clarke (1993), and non-metric multi-dimensional scaling (nmds) ordinations were plotted. One-way analyses of similarities (ANOSIM; Clarke and Green 1988) tested for differences in the composition of assemblages among the 3 spatial scales (plots, sites and locations). Pairwise comparisons tested for differences between groups of replicates when ANOSIM was significant and the Bonferroni procedure was used to control the probability of type I error for these multiple comparisons (α = 0.05). The average Bray Curtis dissimilarity between groups of replicates at each spatial scale was calculated by use of the SIMPER program (Clarke 1993). Dissimilarity of 0% indicates that the groups are identical in composition of species, while 100% indicates total dissimilarity between groups. Only those taxa that were relatively abundant (i.e. covering, on average, at least 5% of the plate) were used in the univariate analyses, whereas all taxa (and bare space ) were included in the multivariate analyses. Because this was a small-scale pilot study, it was expected that the power of many tests for differences would be small. Thus, it was important to calculate the power of any non-significant tests and this was done for univariate analyses with post-hoc power analyses as described in Underwood (1997).

4 Variability in subtidal assemblages of epibiota 431 Results The diversity of assemblages growing on the plates was initially relatively small, but tended to increase over time. At each time of sampling, however, the plates were dominated by only four or five taxa (i.e. these taxa accounted for, on average, over 70% of the cover) although the dominant taxa varied among times of sampling. The results of the univariate analyses showed that there was variability in recruitment at different spatial scales for different taxa and that this variability was not consistent through time (Table 1). When differences were detected at the smallest spatial scale (between plots 10s of metres apart) it was often at one particular site (Site 1, Location 1). Moreover, when these differences occurred, it was often still possible to detect differences at larger spatial scales (Table 1). Analyses for the dominant taxa are presented here to illustrate these results (Table 2; Fig. 2). The cyanobacterium Oscillatoria sp. was very abundant on many plates for the first two times of sampling when there were large differences in its percentage cover between locations (1000s of metres apart) and between plots (10s of metres apart) at one site (Fig. 2a). At 19 weeks, however, the percentage cover of Oscillatoria had decreased substantially (Fig. 2a) and, as a result, differences were no longer detected at any spatial scale (Table 2a). Components of variation showed similar patterns among spatial scales for the 6- and 10-week samples, that is, greatest variability between locations, essentially none between sites and large variability between plots (Table 2a). Spirorbid polychaetes (including species of Neodexiospira, Janua and Pileolaria) were also abundant on some plates. At each of the three times of sampling, differences were detected between sites at Location 2 and percentage cover differed between sites at Location 1 after 6 weeks (i.e. differences at the 100s of metres scale; Fig. 2b). Table 1. Spatial scales at which differences in percentage covers of abundant taxa were detected (α = 0.05) for three times of sampling; summary of univariate results Lo, locations (1000s m apart); Si, sites (100s m apart); Pl, plots (10s m apart). Dash indicates no differences at any scale Taxon 6 weeks 10 weeks 19 weeks Oscillatoria Lo, Pl Lo, Pl - Spirorbids Si Si Si Colpomenia Pl Lo, Pl Lo Cladophorales Si Lo Si Bachelotia - Si Si Ceramiales - Pl Lo Hydroids Pl Pl Si Bryozoans - Si - Bare space Pl Si, Pl - The large apparent differences between locations were not detected by the ANOVA because the power of this test was very small (< 0.4 for each time) and any differences were masked by the larger differences between sites (Table 2b). Components of variation indicated that, for each time of sampling, the variability between locations was far greater than (at least double) that between sites, while there was essentially no variation between plots (Table 2b). Recruitment of the brown alga Colpomenia sinuosa varied significantly among plots at two sites after 6 weeks (Fig. 2c; Table 2c), and there were differences in the cover of Colpomenia between plots and locations by 10 weeks (Fig. 2c). Small-scale differences had disappeared by 19 weeks, but the differences between locations persisted (Table 2c; Fig. 2c). Variance components for each spatial scale tended to be smaller for Colpomenia than for other taxa (Table 2c). The greatest variability occurred between plots after 6 weeks, then between locations and plots, respectively, after 10 weeks, and by 19 weeks there was significant variability only between locations (Table 2c). Filamentous green algae (here grouped together as Cladophorales) were also common on the plates at all three times of sampling, although by 19 weeks these algae were no longer present on plates at one of the locations. Differences between sites and locations occurred at different times of sampling (Table 1). A mat-like ectocarpalean alga (probably Bachelotia antillarum) was initially uncommon and patchily distributed among plates, but had started to cover a large percentage of the plates by 19 weeks. Spatial differences in the cover of Bachelotia were not evident until 10 weeks when sites were found to differ at Location 1 (Table 1). There were also differences between sites after 19 weeks, although these now occurred at Location 2 (Table 1). Hydroids (including Obelia sp. and Campanularia sp.) were very patchy in their distribution after 6 and 10 weeks and there were differences in percentage cover at the scale of plots at these times (Table 1). However, the percentage cover of hydroids had started to increase at Location 1 by 19 weeks when differences between the two sites developed (Table 1). Red filamentous ceramialean algae (including Polysiphonia sp.) were uncommon, but despite this, differences were still found between plots at one site after 10 weeks (Table 1). There was a significant difference in the cover of ceramialean algae between locations by 19 weeks. Bryozoans (primarily Watersipora subtorquata) were also relatively rare on the plates, but did tend to increase slightly over time. There were differences in the cover of bryozoans between sites at one location by the 10-week sample (Table 1). As the percentage cover of various taxa tended to increase over time, there was a corresponding decrease in the percentage cover of bare space (Fig. 2d). Bare space describes areas of the plates where no organisms were

5 432 T. M. Glasby visible under the microscope, but it is quite likely that they were covered by microfilms of bacteria etc. Spatial differences in the cover of bare space were apparent after 6 weeks when there were differences between plots at Site 1, Location 1 and also after 10 weeks when differences between plots and between sites were detected (Table 2d). As expected, the variance components for each spatial scale differed markedly among times of sampling (Table 2d). There was great variability at most spatial scales up to 10 weeks, after which time there was little bare space on the plates and, thus, less variability at each spatial scale (Table 2d). It is important to note that the power of most tests for which no significant difference was detected was generally very small (often less than 0.3). This was highlighted by the fact that differences were occasionally not detected despite the fact that there was considerable variability at a given spatial scale (as indicated by large variance components). Fig. 2. Percentage covers of selected taxa on settlement plates at plots, sites and locations after 6, 10 and 19 weeks. Differently coloured bars represent the two plots at each site.

6 Table 2. Analyses of variance for dominant taxa for the 3 times of sampling (6, 10 and 19 weeks) Post-hoc pooling done at P > 0.25; superscripts a on MS indicate those terms were pooled and the resultant term was used as denominator for the superscripted F-ratio. NS P > 0.05, *P < 0.05, **P < 0.01, ***P < All data were untransformed. Variances were homogeneous at P > 0.05 for all tests except for Oscillatoria 6 and 19 weeks, spirorbids 6 weeks and Colpomenia 10 and 19 weeks. vc, variance component Source 6 weeks 10 weeks 19 weeks df MS F P vc MS F P vc MS F P vc Variability in subtidal assemblages of epibiota (a) Oscillatoria Location a ** a ** a NS 1.5 Site(L) a 0.84 > a 0.47 > a 0.53 > Plot(S(L)) a 4.96 ** a 5.04 ** a 1.31 > Residual a (b) spirorbids Location NS NS NS Site(L) a *** a *** a *** Plot(S(L)) a 0.27 > a 0.39 > a 1.24 > Residual a a a (c) Colpomenia Location a NS a ** a *** 79.6 Site(L) a 0.86 > a 0.61 > a 1.21 > Plot(S(L)) a *** a 3.27 * a 0.35 > Residual a (d) bare space Location a NS NS NS 43.4 Site(L) a 1.48 > * NS 68.2 Plot(S(L)) a 5.31 ** * NS 19.8 Residual

7 434 T. M. Glasby The lack of power was generally attributable to great variability among replicate samples. Tests for differences between plots were often not powerful despite the relatively good degrees of freedom (df) for these comparisons (4, 16 df). Differences in the overall composition of assemblages growing on the plates were observed at all of the spatial scales examined for each time of sampling (Fig. 3). The 3 replicate plates at each plot were pooled for the nmds ordinations, giving 6 replicates at each site at each of two locations. This was done because I was unable to test for differences between individual plots (using pairwise comparisons) at a significance level of less than P = 0.1 with only 3 replicates (when comparing two groups of three numbers, the maximum number of permutations is 10, therefore, each permutation occurs with a probability of 1/10). Although the sample statistic (Global R) from the ANOSIM for the plot comparison with 3 replicates was always significant, indicating that some plots were significantly different, without being able to do pairwise comparisons with α < 0.1, it was not possible to identify, with any confidence, whether those plots that were 10s of metres apart differed. For testing my hypotheses, it would have been important only if the two plots within each site differed (i.e. those plots that were 10s of metres apart, rather than plots from different sites or locations which would be 100s or 1000s of metres apart). The nmds ordinations indicated that, for the first two times of sampling, there was a lot of variability among individual plates, particularly at Location 1, Site 1 (Fig. 3a, b). There appeared to be a difference between the two plots at this site by 19 weeks (i.e. 2 groups of 3 points; Fig. 3c), but there were no other distinct separations of plots within sites (Fig. 3). Sites and locations were clearly separated, particularly after 10 and 19 weeks (Fig. 3b, c). ANOSIM and pairwise comparisons showed that, at each time of sampling, there was a significant difference between the two locations and between the sites at Location 1. Thus, even if there were differences between plots, differences between sites and locations could still be detected. This was supported by the comparison of average dissimilarities among the three spatial scales. The dissimilarity between plots nested within sites was consistently less than that between sites nested within locations which, in turn, was less than that between locations (Table 3). Moreover, the dissimilarities between plots and between sites tended to decrease over time as the assemblages became more developed, while the dissimilarity between locations was consistently large (Table 3). Table 3. Average percent dissimilarity (± s.e.m.) between plots, sites and locations for the three times of sampling There are no standard errors for the comparison between locations, because there was only ever one of these comparisons per time of sampling (there were 4 for plots and 2 for sites) Spatial scale 6 weeks 10 weeks 19 weeks Plots 28.3 (± 6.6) 24.6 (± 2.4) 17.9 (± 3.0) Sites 33.4 (± 7.1) 30.6 (± 0.9) 25.9 (± 4.1) Locations Discussion This study showed that there is variability in developing epibiotic assemblages around Sydney, Australia over spatial scales of 10s to 1000s of metres. The greatest variability was between sites (100s of metres apart) and locations (1000s of metres apart). When plots were found to be different, it was generally still possible to detect differences at larger spatial scales (i.e. between sites or locations). These results are consistent with the findings of the few other studies that Fig. 3. nmds ordinations for the 3 times of sampling, comparing percentage covers of taxa at Location 1 (circles), Location 2 (triangles) and the two sites nested within locations (1, filled symbols; 2, open symbols). Plots at each site have not been distinguished, thus n = 6 per site.

8 Variability in subtidal assemblages of epibiota 435 have examined the variability in recruitment of subtidal organisms onto hard substrata. Patterns of recruitment of subtidal invertebrates have been found to vary over scales of 10s of metres to 10s of kilometres (Kay and Keough 1981; Keough 1983; Butler 1986). Moreover, patterns of larval recruitment at different spatial scales are unlikely to be the similar for different taxa (Butler 1986). The scale at which the percentage covers of taxa varied was not consistent through time (i.e. at different stages of development of the assemblage and/or with time of year). This was especially evident from the comparison of variance components. Rarely were there similar degrees of variability among the spatial scales for each time of sampling, nor did the cover of a particular taxon regularly vary at the same spatial scales. This was most likely due to a combination of variability in recruitment of species over time (e.g. Sutherland and Karlson 1977; Osman 1977, 1978; Keough 1983) and changes in the structure of the assemblages over time (e.g. the disappearance of Oscillatoria), which may, in itself, affect further recruitment (Osman and Whitlatch 1995). The abundance of organisms such as the cyanobacterium Oscillatoria and filamentous green algae varied greatly over time. When they were relatively abundant, there were clear differences in the percentage covers of these organisms over a variety of spatial scales. Conversely, when these taxa were uncommon on plates, no significant differences were detected at any spatial scale. Despite the changing abundances of individual taxa over time, the multivariate analyses showed that there was always variability in the overall composition of the assemblages at scales of 100s and 1000s of metres. Most of the non-significant univariate tests had very small power (i.e. < 0.3) so it is possible that differences were not detected because of inadequate replication. This is not surprising given the limited nature of the study, but clearly it is possible that significant differences could have been detected if greater replication was used. That is, this study has most likely under-estimated the scales of spatial variability in recruitment of subtidal sessile organisms. Nevertheless, the results do demonstrate that variability may occur at a variety of spatial scales and that the exact nature of this variability may change as assemblages of epibiota develop. The multivariate analyses indicated that variability at the scales of 10s and 100s of metres tended to decrease as the assemblages developed, whereas variability between locations 1000s of metres apart continued to be great. The spatial variability demonstrated in this and other studies must be estimated as part of any environmental impact study involving epibiota on subtidal hard substrata. For comparisons among areas 100s or 1000s of metres apart, replicates within each of these areas will need to be spaced at least 10s of metres apart to incorporate small-scale variability. Moreover, it is likely that considerable replication will be necessary to ensure that powerful tests can be constructed. Inadequate small-scale spatial replication is likely to confound comparisons among locations (Underwood 1981, 1989, 1993; Hurlbert 1984; Andrew and Mapstone 1987; Morrisey and Underwood 1992). Furthermore, a lack of understanding about largescale variability might result in the choice of inappropriate control or reference locations (Green 1979; Underwood 1992, 1994; Keough and Mapstone 1995). Without detailed knowledge of scales of natural variability it will be impossible to make any sensible conclusions about environmental impacts (Lewis 1976; Christie 1980, 1985). Settlement plates are seldom made from natural substrata as was done here (but see Kennelly 1983; Hixon and Brostoff 1985; McGuinness 1989). Use of a natural substratum reduces the likelihood that organisms will respond in different ways to the settlement plates and natural surfaces. Of course, settlement plates may often be unlike natural surfaces in that they are small, isolated patches (Keough 1984a; Connell and Keough 1985; Butler 1991). Nevertheless, attachment and growth of organisms should be similar on natural surfaces and settlement plates made from the same natural substratum. Thus, settlement plates made from a naturally occurring substratum will be useful for relative comparisons among locations, but may not necessarily provide specific information about assemblages growing on large surfaces such as rocky reefs. The design and deployment of the settlement plates used here helped overcome potential problems of other studies that use settlement plates. First, any effects of supporting structures (such as frames, screws etc.) on the settlement and growth of organisms were minimized by the use of plates that were essentially free-standing in the water column. The area of the plate sampled was isolated from surfaces other than sandstone. Second, the estimates of the composition of assemblages at different stages of development were independent of one another, but were directly comparable because the plates were all deployed simultaneously and collected after varying amounts of time. This has only rarely been done for studies of developing assemblages of epibiota (e.g. Shin 1981; Underwood and Anderson 1994). Finally, the nested sampling design was useful for identifying the scales of spatial variability in recruitment of subtidal sessile epibiota. The scales of natural variation in recruitment may not always be obvious, in which case this kind of hierarchical design should be used to determine the appropriate scales at which to sample (e.g. Fairweather 1990; Jones et al. 1990; Morrisey et al. 1992; Thrush et al. 1994). Pilot studies are often time-consuming, but they are frequently an important beginning to ecological research (Green 1979; Underwood 1981; Andrew and Mapstone 1987). The data presented here will be useful for the design

9 436 T. M. Glasby of future studies on spatial variability in recruitment of subtidal epibiota and also for the planning of studies of environmental impact in this habitat. This type of information is vital to any test for environmental impact. Without it, inappropriate and possibly confounded comparisons among locations are likely. Acknowledgments This work was facilitated by an Australian Postgraduate Award and funds from the Institute of Marine Ecology and, during the preparation of the manuscript, the Centre for Research on Ecological Impacts of Coastal Cities (University of Sydney). Professor Tony Underwood gave considerable advice throughout all stages of this research. I thank V. Mathews, G. Housefield, B. Gillanders, W. McBeth, P. Higgs, M. Bajenov and A. Jinks for assistance in the field. Dr G. Rouse identified polychaetes and Prof. Tony Larkum identified algae. Sandstone was kindly provided by Gosford Quarries Pty Ltd. Earlier versions of the manuscript benefited from comments by M. W. Beck, A. J. Butler, J. S. Gray, C. D. Todd, A. J. Underwood and three anonymous reviewers. References Andrew, N. L., and Mapstone, B. D. (1987). Sampling and the description of spatial pattern in marine ecology. Annual Review of Oceanography and Marine Biology 25, Ayling, A. M. (1981). The role of biological disturbance in temperate subtidal encrusting communities. Ecology 62, Bray, J. R., and Curtis, J. T. (1957). An ordination of the upland forest communities of southern Wisconsin. Ecological Monographs 27, Butler, A. J. (1986). Recruitment of sessile invertebrates at five sites in Gulf St. Vincent, South Australia. Journal of Experimental Marine Biology and Ecology 97, Butler, A. J. (1991). Effect of patch size on communities of sessile invertebrates in Gulf St. Vincent, South Australia. Journal of Experimental Marine Biology and Ecology 153, Butler, A. J., and Chesson, P. L. (1990). Ecology of sessile animals on sublittoral hard substrata: The need to measure variation. Australian Journal of Ecology 15, Caley, M. J., Carr, M. H., Hixon, M. A., Hughes, T. P., Jones, G. P., and Menge, B. A. (1996). Recruitment and the local dynamics of open marine populations. Annual Review of Ecology and Systematics 27, Christie, H. (1980). Methods for ecological monitoring: biological interactions in a rocky subtidal community. Helgoländer Meeresuntersuchungen 33, Christie, H. (1985). Ecological monitoring strategy with special reference to a rocky subtidal programme. Marine Pollution Bulletin 16, Clarke, K. R. (1993). Non-parametric multivariate analyses of changes in community structure. Australian Journal of Ecology 18, Clarke, K. R., and Green, R. H. (1988). Statistical design and analysis for a biological effects study. Marine Ecology Progress Series 46, Connell, J. H., and Keough, M. J. (1985). Disturbance and patch dynamics of subtidal marine animals on hard substrata. In The ecology of natural disturbance and patch dynamics. (Eds S. T. A. Pickett and P. S. White.) pp (Academic Press: New York.) Dean, T. A., and Hurd, L. E. (1980). Development in an estuarine fouling community: the influence of early colonists on later arrivals. Oecologia (Berlin) 46, Fairweather, P. G. (1990). Sewage and the biota on seashores: assessment of impact in relation to natural variability. Environmental Monitoring and Assessment 14, Foster, M. S. (1975a). Algal succession in a Macrocystis pyrifera forest. Marine Biology 32, Foster, M. S. (1975b). Regulation of algal community development in a Macrocystis pyrifera forest. Marine Biology 32, Green, R. H. (1979). Sampling design and statistical methods for environmental biologists. (Wiley Interscience: Chichester, England.) Hirata, T. (1987). Succession of sessile organisms on experimental plates immersed in Nabeta Bay, Izu Peninsula, Japan. II. Succession of invertebrates. Marine Ecology Progress Series 38, Hixon, M. A., and Brostoff, W. N. (1985). Substrate characteristics, fish grazing, and epibenthic reef assemblages off Hawaii. Bulletin of Marine Science 37, Hurlbert, S. J. (1984). Pseudoreplication and the design of ecological field experiments. Ecological Monographs 54, Jones, G. P., Ferrell, D. J., and Sale, P. F. (1990). Spatial pattern in the abundance and structure of mollusc populations in the soft sediments of a coral reef lagoon. Marine Ecology Progress Series 62, Kay, A. M., and Butler, A. J. (1983). Stability of the fouling communities on the pilings of two piers in South Australia. Oecologia (Berlin) 56, Kay, A. M., and Keough, M. J. (1981). Occupation of patches in the epifaunal communities on pier pilings and the bivalve Pinna bicolour at Edithburgh, South Australia. Oecologia (Berlin) 48, Kennelly, S. J. (1983). An experimental approach to the study of factors affecting algal colonization in a sublittoral kelp forest. Journal of Experimental Marine Biology and Ecology 68, Kennelly, S. J., and Underwood, A. J. (1993). Geographic consistencies of effects of experimental physical disturbance on understorey species in sublittoral kelp forests in central New South Wales. Journal of Experimental Marine Biology and Ecology 168, Keough, M. J. (1983). Patterns of recruitment of sessile invertebrates in two subtidal habitats. Journal of Experimental Marine Biology and Ecology 66, Keough, M. J. (1984a). Effects of patch size on the abundance of sessile marine invertebrates. Ecology 65, Keough, M. J. (1984b). Dynamics of the epifauna of the bivalve Pinna bicolor: interactions among recruitment, predation, and competition. Ecology 65, Keough, M. J., and Butler, A. J. (1995). Temperate subtidal hard substrata. In State of the Marine Environment Report for Australia. Technical Annex 1. The Marine Environment. (Eds L. P. Zann and P. Kailola.) pp (Great Barrier Reef Marine Park Authority: Townsville.) Keough, M. J., and Mapstone, B. D. (1995). Protocols for designing marine ecological monitoring programs associated with BEK mills. National Pulp Mills Research Program, Technical Report No. 11 (CSIRO: Canberra.) 185 pp. Lewis, J. R. (1976). Long-term ecological surveillance: practical realities in the rocky littoral. Annual Review of Oceanography and Marine Biology 14, McGuinness, K. A. (1989). Effects of some natural and artificial substrata on sessile marine organisms at Goleta Reef, Panama. Marine Ecology Progress Series 52, Morrisey, D. J., and Underwood, A. J. (1992). Sampling for spatial variation in the distribution of fauna in sediments. In Proceedings of a Bioaccumulation Workshop: Assessment of the Distribution, Impacts and Bioaccumulation of Contaminants in Aquatic Environments. (Ed. A. G. Miskiewicz.) pp (Water Board and Australian Marine Sciences Association: Sydney.)

10 Variability in subtidal assemblages of epibiota 437 Morrisey, D. J., Howitt, L., Underwood, A. J., and Stark, J. S. (1992). Spatial variation in soft-sediment benthos. Marine Ecology Progress Series 81, Osman, R. W. (1977). The establishment and development of a marine epifaunal community. Ecological Monographs 47, Osman, R. W. (1978). The influence of seasonality and stability on the species equilibrium. Ecology 59, Osman, R. W., and Whitlatch, R. B. (1995). The influence of resident adults on recruitment: a comparison to settlement. Journal of Experimental Marine Biology and Ecology 190, Russ, G. R. (1980). Effects of predation by fishes, competition, and the structural complexity of the substratum on the establishment of a marine epifaunal community. Journal of Experimental Marine Biology and Ecology 42, Russ, G. R. (1982). Overgrowth in a marine epifaunal community: competitive hierarchies and competitive networks. Oecologia (Berlin) 53, Shin, P. K. S. (1981). The development of sessile epifaunal communities in Kylesalia, Kilkieran Bay (west coast of Ireland). Journal of Experimental Marine Biology and Ecology 54, Sutherland, J. P., and Karlson, R. H. (1977). Development and stability of the fouling community at Beaufort, North Carolina. Ecological Monographs 47, Thrush, S. F., Pridmore, R. D., and Hewitt, J. E. (1994). Impacts on softsediment macrofauna: the effects of spatial variation on temporal trends. Ecological Applications 4, Todd, C. D., and Turner, S. J. (1986). Ecology of intertidal and sublittoral cryptic epifaunal assemblages. I. Experimental rationale and the analysis of larval settlement. Journal of Experimental Marine Biology and Ecology 99, Underwood, A. J. (1981). Techniques of analysis of variance in experimental marine biology and ecology. Annual Review of Oceanography and Marine Biology 19, Underwood, A. J. (1989). The analysis of stress in natural populations. Biological Journal of the Linnean Society 37, Underwood, A. J. (1991). Beyond BACI: experimental designs for detecting human environmental impacts on temporal variations in natural populations. Australian Journal of Marine and Freshwater Research 42, Underwood, A. J. (1992). Beyond BACI: the detection of environmental impact on populations in the real, but variable, world. Journal of Experimental Marine Biology and Ecology 161, Underwood, A. J. (1993). The mechanics of spatially replicated sampling programmes to detect environmental impacts in a variable world. Australian Journal of Ecology 18, Underwood, A. J. (1994). On beyond BACI: sampling designs that might reliably detect environmental disturbances. Ecological Applications 4, Underwood, A. J. (1997). Experiments in ecology: their logical design and interpretation using analysis of variance. (Cambridge University Press: Cambridge.) Underwood, A. J., and Anderson, M. J. (1994). Seasonal and temporal aspects of recruitment and succession in an intertidal estuarine fouling assemblage. Journal of the Marine Biological Association of the United Kingdom 74, Underwood, A. J., and Denley, E. J. (1984). Paradigms, explanations and generalizations in models for the structure of intertidal communities on rocky shores. (Princeton University Press: New Jersey.) Winer, B. J. (1971). Statistical Principles in Experimental Design, 2nd Edn. (McGraw Hill Kogakusha: Tokyo.) Manuscript received 19 January 1998; revised and accepted 5 June 1998