Community assembly in macrofauna communities developed in artificial kelp holdfasts was monitored at 1-month intervals over a 13 month period using a sampling design that used systematic patterns of temporal overlap and changes in start and collection dates. The hierarchical nature of the experimental design allowed several different approaches to analysis; by date of deployment and by date of collection of the artificial habitats, which enabled comparison of community assembly with and without the seasonal effects of the date of collection, and by community age to test whether there were alternative end-states to assembly depending on season or recruitment history.
Artificial kelp holdfasts were constructed of bundles of 10, 150-mm lengths of polypropylene rope bound together at one end with a plastic cable tie. At the other end, the rope strands were separated, splayed and glued to a 100 mm x 100 mm PVC base. The design was similar in size and physical complexity to the holdfasts of Ecklonia radiata, the most common species of kelp in southern Australia.
Holdfasts were attached to concrete bricks with cable ties and distributed over sand adjacent to a healthy E. radiata dominated reef. The study site was 100 m x 4 m, and holdfasts were deployed randomly to 1 m grid co-ordinates. The rocky reef ran the full length (100 m) of the grid and was likely the major source of recruits to the holdfasts.
Holdfasts were deployed each month for 13 months beginning in December 1997. At each deployment, sufficient holdfasts were established to collect 6 replicate holdfasts each subsequent month until January 1999 (Figure 2). An earlier pilot study indicated that 6 holdfasts was the minimum sampling intensity to adequately estimate natural variability across replicate holdfast communities, standard errors in the abundance of individual taxa did not significantly decrease with a greater number of replicates. Deployment and collection dates were toward the end of the nominated month, weather permitting. Water temperature peaked in January and February at 18oC and reached a minimum of 11oC during June, July and August. A total of 408 holdfasts were deployed to random positions on the grid and later recovered.
Monthly collections involved recovering 6 randomly selected replicate holdfasts from each previous month of deployment. Holdfasts were gently covered with a plastic bag before cutting the cable ties attaching the holdfast to its concrete brick and sealing the bag for transport to the surface. Vacated grid-positions were open to subsequent deployment of another artificial holdfast (if randomly selected). Due to poor weather, holdfasts could not be collected or deployed in September while collections were not possible in June. For treatments deployed in December 1997, 4 replicates (rather than 6) were collected each subsequent month.
Holdfasts were preserved in ~5% buffered formalin. For processing, holdfasts were readily broken open (by the cutting of the cable tie) and washed thoroughly over a 1 mm sieve to remove all animals. Solitary animals retained on the sieve were identified where possible to the level of family, the most notable exception being amphipods, which were identified to sub-order. Colonial organisms were not enumerated because they were very rare. We deemed taxonomic resolution to the level of family as the optimal cost-benefit trade-off given the large abundance of organisms encountered (148, 841 individuals), and that family-level patterns typically reflect patterns at the species-level