Griffiths et al.: Fish recolonization in temperate Australian rockpools 



637 



by chance. Similarity percentages (SIMPER) were used 

 to determine which species were responsible for differ- 

 ences between selected groups. This analysis involved 

 calculating the average contribution of each species in 

 each pair of groups and comparing this contribution to 

 the overall dissimilarity of fish assemblages between 

 the groups. All multivariate analyses were carried out 

 with PRIMER (Plymouth routines in multivariate eco- 

 logical research) software (version 5.2.2, PRIMER-E 

 Ltd., Roborough, Plymouth, UK). 



Results 



Composition of rockpool fish assemblages 



A total of 3658 fish representing 38 species and 19 

 families was caught in 176 samples from 32 rockpools at 

 Bass Point between 7 September 1999 and 22 September 

 2000 (Table 1), corresponding to densities of 0.5 and 19 

 species/m 3 (mean 4.4 [±2.9] /m 3 ) and 0.5 and 80 fish/m 3 

 (mean 15.6 | ±14.6] /m 3 ), respectively. The most numeri- 

 cally abundant taxa were permanent rockpool residents 

 representing the families Gobiidae iBathygobius cocosen- 

 sis), Tripterygiidae iEnneapterygius rufopileus), Clinidae 

 iHeteroclinus whiteleggi and H. fasciatus), Blenniidae 

 (Pa?-ablenrjius intermedins), and Gobiesocidae {Aspasmo- 

 gaster costatus), although the temporary resident Girella 

 elevata was the third most abundant species. The ten 

 most numerically abundant species represented 92% of 

 the catch (Table 1). Three species, G. elevata, Scorpis 

 lineolatus, and Myxus elongatus, represented by 504 

 fish were considered to be of economic significance. All 

 economically important fishes were caught as juveniles 

 in the rockpools and 89% of the fish measured less than 

 100 mm FL. 



Numbers of species and individuals 



For the short-term studies, the mean number of species 

 differed significantly between sampling intervals and 

 sites (RM-ANOVA, Table 2). With respect to the site 

 factor, there were significantly more species caught at 

 BB than at the other three sites and the latter three 

 sites did not differ from each other (SNK test). Only 

 the "1-week" samples accounted for significantly fewer 

 species than the initial samples (Fig. 2). However, the 

 mean number of species caught in the "1-month" and 

 "3-month" samples did not differ significantly from the 

 initial samples at all sites (Fig. 2). 



The mean number of individuals differed significantly 

 between sampling intervals and sites, although there 

 was also a significant time x site interaction (RM-ANO- 

 VA, Table 2). A close investigation of the significant 

 interval xsite interaction, with primary interest in the 

 interval factor, revealed that the number of individuals 

 in the initial samples did not differ significantly from 

 samples taken after three months at the exposed sites 

 (MB and TC), but they did differ significantly at shel- 

 tered locations (GL and BB) (Fig. 2). It appeared that the 



Number of species 



£ 50 



-5 30 _ 



I 10 



Number of individuals 



; ni1^inW 



GL 



Site 



Figure 2 



Mean i±SE) numbers of species and individuals (m~ 3 ) 

 caught in rockpools at Bass Point, New South Wales 

 between 7 September 1999 and 17 September 2000 

 during the short-term recolonization studies (combined 

 for spring and autumn) between sampling intervals 

 separated by 1 week, 1 month, and 3 months. Key to 

 sites: Maloney's Bay (MB). The Chair (TC I, Gravel Loader 

 (GL), Beaky Bay <BBi. 



Loader and Beaky Bay sites initially supported unusu- 

 ally high numbers of individuals and these high numbers 

 may have accounted for significantly fewer individuals 

 caught in the subsequent samples (Fig. 2). 



For the long-term study, the number of individuals 

 significantly differed among sampling times but did 

 not for number of species I RM-ANOVA, Table 3). The 

 significant difference in the mean number of individuals 

 was due to fewer individuals caught in the "12-month" 

 samples when fish numbers were pooled for all sites 

 (SNK test, Fig. 3). 



Variation in abundance of major recolonizing species 



The rank abundances of the numerically dominant 

 species were consistent for B. cocosensis and E. rufopi- 

 leus across all sampling intervals for all studies, even 

 though their relative abundances varied considerably 

 (Table 4). In contrast, the ranks of the least common 

 of the six species, namely H. whiteleggi, P. intermedius. 



