Materials and methods Directory UMM :Data Elmu:jurnal:J-a:Journal of Experimental Marine Biology and Ecology:Vol248.Issue2.May2000:

T .M. Glasby J. Exp. Mar. Biol. Ecol. 248 2000 177 –190 179 naturally occurring rocky overhangs were not made because of the rarity and small size of overhangs in local estuaries. Nevertheless, assemblages on the undersides of pontoons appear to be similar to those under rocky overhangs and not as different as assemblages on vertical surfaces of the two structures personal observation. It was proposed, therefore, that factors associated only with undersides of surfaces perhaps increased shading, different patterns of water flow, etc. can over-ride at least partially or obscure any differences that occur between assemblages on vertical surfaces of natural and artificial structures. Thus, different assemblages were predicted to develop on vertical surfaces of different compositions, but not on horizontal undersides. The differences among vertical sandstone, concrete and wooden surfaces were predicted to be similar to those described previously Connell and Glasby, 1999 among sandstone reef and wooden and concrete urban structures.

2. Materials and methods

Settlement panels were deployed at three sites, each | 200 m apart, in Middle Harbour 338489 S, 1518149 E, the northern part of Sydney Harbour, in June winter 1998. Middle Harbour is a sheltered waterway | 5 km from the open ocean and a residential area. At each site, the panels were on reef adjacent to pontoons and jetties, but not directly shaded by them. Settlement panels were 15 3 15 cm and constructed of either sandstone, concrete or marine plywood. These materials were chosen to represent the more common hard substrata in Sydney waterways — reefs and retaining walls are primarily sandstone, pilings are constructed of wood or concrete and many pontoons are concrete. Pilings are constructed of a variety of different timbers mainly hardwood, some of which are treated with antifouling paints. The marine plywood used in the present study was not treated with antifouling paint and may have had different surface characteristics from many wooden pilings. Aluminium beams 32 3 32 mm aluminium 908 angle were drilled onto concrete blocks on the rocky reef at each site at a depth of 1.5 m below mean low water springs. PVC brackets were glued on to the backs of panels and these were bolted to the beams as in Glasby, 1998. Three of the beams at each site supported vertical panels, three supported horizontal panels facing downwards. The six beams were arranged in three pairs, such that vertical and horizontal panels were within 50 cm of each other. There were five panels on a beam and either one or two replicates of each surface were attached to a beam. Replicates on the same beam were | 1 m apart and replicates on different beams were | 5 m apart. There were five replicates of each surface, of each orientation per site. All panels were 10–15 cm from the substratum, but were still accessible to benthic grazers such as urchins and gastropods. Panels were collected in summer after 7 months, suspended and supported in tubs of seawater for transport back to the laboratory where they were transferred into filtered 10 mm filter seawater and refrigerated at 58C until sorted within 6 days of collection under a dissecting microscope. Primary cover organisms attached directly to the panel and secondary cover organisms attached to primary cover were estimated for sessile organisms on the fronts of panels using 64 regularly spaced points in a 13 3 13 cm grid 180 T .M. Glasby J. Exp. Mar. Biol. Ecol. 248 2000 177 –190 i.e. a 1 cm border around each panel was not sampled to avoid ‘edge effects’. Taxa on the fronts of panels, but not under a point were assigned a cover of 0.5. Species of filamentous algae were lumped together because often many species were present under a point and it was impossible to allocate the point to just one species. Data for primary and secondary covers of each taxon were combined and expressed as a percentage for univariate ANOVA and non-parametric multivariate PRIMER package analyses. Variances were heterogeneous for many analyses Cochran’s C-test, P . 0.05, despite transformation, so untransformed data were analysed. For balanced designs, ANOVA is relatively robust to heterogeneous variances Box, 1953, but significant results at the level of P , 0.05 should perhaps be interpreted with some caution. Multivariate data were fourth root transformed and Bray–Curtis similarity matrices calculated. Data were presented graphically using non-metric multi-dimensional scaling nMDS ordinations. One-way analyses of similarities ANOSIM; Clarke and Green, 1988 and multiple pairwise comparisons tested for differences in assemblages among treatments. The significance level for pairwise tests was reduced from 0.05 to 0.01 to adjust for multiple comparisons. Some panels were lost from Site 1 one sandstone, two concrete and two wood and to maintain balanced univariate analyses, these values were replaced by the mean of the remaining replicates at that site and the residual degrees of freedom, mean square estimates and F-ratios adjusted accordingly for each taxon Underwood, 1981.

3. Results