644 



Fishery Bulletin 102(4) 



Spring 



Autumn 



/~> Long-term 



Initial n=73 



i — i i F 1 i — I 



10 20 30 40 50 60 



Total length (mm) 



Figure 7 



Length-frequency distributions (in 2-mm intervals) for Enneapterygius rufopileus 

 caught in the initial samples and after 1 week, 1 month, and 3 months later in the 

 short-term study and caught in the initial sampling and 6 months and 12 months 

 later in the long-term study. Samples were pooled for all sites during the (A) spring 

 and (B) autumn short-term recolonization studies and for all sites during (C) the 

 long-term recolonization study. Studies were conducted between 7 September 1999 

 and 22 September 2000. Samples sizes are shown. 



because samples were generally widely dispersed in 

 nMDS ordination plots. The relatively low stress values 

 (<0.2) indicate that high variability in fish assemblages 

 at the level of individual rockpools is probably respon- 

 sible for the patterns observed. However, some spatial 

 variability in fish recolonization patterns was evident 

 (Fig. 2) and appeared to be dependent to some extent 

 on exposure of sites to predominant swell. There is 

 evidence to suggest that wave exposure can affect the 

 distribution of intertidal fishes (Gibson, 1972; Ibanez et 

 al., 1989), although there is apparently no study that 

 has investigated this effect in relation to fish recoloni- 

 zation in rockpools. In the present study, recolonization 

 appeared more rapid at wave-exposed sites (MB and 

 TC) compared to more sheltered sites. This may have 

 been the result of the close distance between rockpools 

 at exposed sites (within meters of each other), whereas 

 at both sheltered sites, rockpools were significantly 

 farther apart. Consequently, defaunated rockpools at 

 exposed sites may recolonize more quickly if the major 

 recolonizers are derived from neighboring rockpools 

 as has been documented elsewhere (Beckley, 1985a; 

 Polivka and Chotkowski, 1998). 



Fish recolonization patterns were not influenced by 

 the time of year that rockpools were defaunated in 

 either short-term or long-term studies. The numbers 



of species and individuals consistently returned to pre" 

 perturbation levels within a few weeks, but this return 

 to previous levels may partially be a consequence of 

 the relatively small number of species that are nor- 

 mally found in rockpools at any given time. In such 

 situations a significant differences could only occur if 

 large-scale changes in abundances were recorded. The 

 lack of temporal variation in recolonization rates was 

 surprising because recolonization was expected to be 

 more rapid during summer, when the larvae of residents 

 and warm water transients are expected to be avail- 

 able for settlement (Beckley, 1985a; Willis and Roberts, 

 1996). Recruitment was not the major mechanism driv- 

 ing fish recolonization in the present study because the 

 majority of recolonizers were subadults and adults that 

 would have relocated from nearby rockpools. Although 

 many of the fish captured in each pool were tagged, 

 the vast majority offish caught in the same rockpool in 

 subsequent sampling events were not tagged. Griffiths 

 (2003b) showed that the common recolonizing species in 

 the present study moved between a few rockpools within 

 a limited home range. Therefore, postsettlement fishes 

 from surrounding rockpools were probably moving into 

 the study rockpools between each sampling event. 



The movement of postsettlement fishes from adjacent 

 rockpools also appears to control the resilience of rock- 



