640 



Fishery Bulletin 102(4) 



est in the 1-week samples at each site for the majority 

 of dominant species. For the long-term study only the 

 mean number of B. cocosensis and E. rufopileus differed 

 significantly among sampling times (Table 5), and this 



E 30 



ro 25 



Number of species 



35 "I Number of individuals 



Figure 3 



Mean (±SE) numbers of species and individuals i m'i 

 caught in rockpools at Bass Point, New South Wales, 

 between 22 September 1999 and 11 September 2000 

 (pooled for all four sites) during the long-term recoloniza- 

 t ion study. Intervals between sampling for the long-term 

 study were six months. 



difference was due to lower numbers being caught in 

 the 12-month samples (Fig. 4). 



Fish assemblage structure 



No clear patterns emerged in the nMDS ordination plots, 

 with the exception of separation of the "1 week" samples 

 from all other samples at BB during the autumn study 

 (Fig. 5). ANOSIM supported these visual interpretations 

 of ordination plots and revealed that fish assemblages 

 did not differ significantly among sampling times at any 

 of the four sites for the spring and long-term studies 

 (Table 6) because abundant species E. rufopileus, B. coco- 

 sensis, H. fasciatus, and P. intermedius were common in 

 all samples (SIMPER analysis). For the autumn study, 

 the results of ANOSIM complemented those of RM- 

 ANOVA in that significant differences among sampling 

 intervals were detected only at BB (Table 6). At this 

 site the initial samples and "1-week" samples differed 

 significantly in their fish assemblages, which was due 

 to higher numbers of G. elevata and B. cocosensis in the 

 initial samples (SIMPER analysis). 



Length-frequency distributions 



Removal of fishes from rockpools did not have any 

 apparent effects on the length-frequency distributions 

 for at least two species (B. cocosensis and E. rufopi- 

 leus) for which there were sufficient data to construct 

 length-frequency histograms. Unfortunately, the less 

 abundant recolonizing species, namely P. intermedius, 

 A. costatus, and H. whiteleggi, were caught in too few 

 numbers to ascertain the impacts of defaunation on 

 their size compositions. Rockpools were mainly recolo- 

 nized by subadults and adults for B. cocosensis and E. 

 rufopileus in all three studies (Figs. 6 and 7). However, 

 cohorts of small juveniles (15-30 mm) were evident in 

 the "3-month" samples during spring and the initial 

 autumn studies (February to June), which could then 

 be clearly identified in subsequent samples (Figs. 6 

 and 7). 



