636 



Fishery Bulletin 102(4) 



months in duration) were undertaken in spring-summer 

 and autumn-winter (hereafter referred to as spring and 

 autumn studies, respectively), and a long-term recolo- 

 nization study spanned a 12-month period. Rockpools 

 for each of the three studies were selected at four sites 

 at Bass Point, NSW, which are named Maloney's Bay 

 (MB), The Chair (TO, Gravel Loader (GL), and Beaky 

 Bay (BB) (Fig. 1). Each of the four sites are separated 

 by about 1 km. Rockpools were selected at each site 

 (50-200 m apart) according to similar physical param- 

 eters (i.e., volume, surface area, and substrate type) 

 and particularly according to their vertical elevation 

 on the rock platform. Because higher pools might have 

 less chance of fish recolonization because they are less 

 frequently inundated by seawater (Griffiths et al., 2003). 

 every effort was made to select pools located in the mid- 

 intertidal zone (1-1.5 m above MLLW [mean lower low 

 water]) and, although pools were visually similar, they 

 varied in volume, ranging from 762 to 2160 liters (or 

 0.76-2.16 m 3 ). The bottom of the rockpools consisted of 

 pebbles, cobbles, and small boulders. 



For the short-term studies, four rockpools were sam- 

 pled and fish removed at each of the four sites. In the 

 spring study (beginning 7 September 1999), they were 

 then resampled 1 week, 1 month, and 3 months af- 

 ter the preceding sampling date (referred to as the 

 "1-week," "1-month," and "3-month" samples in this 

 article). This study ended on 8 February 2000, after a 

 period of 5 months. After this date a period of at least 

 three months was given for pools to re-establish fish 

 assemblages before beginning the autumn study on 15 

 May 2000. Rockpools were sampled in exactly the same 

 manner as for the spring study, with sampling ending 

 on 17 September 2000. For each study, 64 samples were 

 taken giving a total of 128 samples for the short-term 

 studies. It is important to note that although every ef- 

 fort was made to resample pools after exactly the same 

 time intervals, this was not possible because of daily 

 time and height of tides and wave heights. For example, 

 for the "1 week" samples, the number of days between 

 samples was actually between 7 and 10 days. 



To determine whether frequent sampling in the short- 

 term studies affected the structure of rockpool fish as- 

 semblages, a long-term study was undertaken by using 

 four different rockpools at the same four sites that were 

 sampled in the short-term studies. Four rockpools at 

 each site were considered adequate because Griffiths 

 (2003a) was able to detect significant differences in the 

 numbers of fish species and individuals in rockpools 

 between sites and months using four rockpools per site 

 in the same region that was surveyed in our study. 

 Rockpools were initially sampled on 22 September 1999 

 and then resampled twice at intervals of six months 

 (20 April 2000 and 11 September 2000). A total of 48 

 samples were taken for this study. 



Data collection 



Fish were collected by hand after completely emptying 

 each rockpool with a VMC 12V battery-powered bilge 



pump of 9029 L/h capacity by using the methods of 

 Griffiths (2000). A thorough search of each pool was 

 conducted by overturning all rocks and boulders, search- 

 ing all crevices and shaking algal fronds until all fish 

 that could be seen were removed. Fishes were identified 

 and total lengths (TL) were measured. Fork length (FL) 

 was also measured for economically significant species. 

 Fish were categorized as being juveniles (<30 mm), 

 subadults (30-40 mm), or adults (>40 mm). Fish were 

 then released alive into rockpools or the shallow subtidal 

 10-30 m away from the rockpool being sampled, which 

 was considered to be the approximate distance that fish 

 may be displaced by waves and surge during significant 

 natural disturbances, such as storms. Each species was 

 categorized by its residential status in rockpool habitats 

 according to the definitions of Griffiths (2003c) in order 

 to better understand the types of fish responsible for 

 recolonization. These categories were "permanent resi- 

 dents," "opportunists," and "transients." 



Statistical analyses 



A repeated-measures ANOVA (RM-ANOVA) was used 

 (SPSS vers. 6.1; SPSS, Chicago, IL) to test for sig- 

 nificant differences in the numbers of species and indi- 

 viduals between sampling intervals (within-subjects 

 factor) and sites (among-subjects factor). Short- and 

 long-term experiments were analyzed with two sepa- 

 rate RM-ANOVAs. For the short-term study a third 

 factor of season (i.e., spring or autumn; among-sub- 

 jects factor) was added. All factors were considered 

 fixed. Assumption of sphericity of the variance-covari- 

 ance matrix was tested by using Mauchly's criterion 

 and, if violated, F tests were performed with Green- 

 house-Geisser-adjusted degrees of freedom. Student- 

 Newman-Keuls iSNK) tests were used for a posteriori 

 comparisons among means (numbers of species and 

 individuals) in RM-ANOVAs. 



Nonmetric multidimensional scaling (nMDS) was 

 used to examine similarities in fish assemblage struc- 

 ture between sampling intervals and sites. Data were 

 fourth-root transformed, to reduce the influence of 

 highly abundant taxa, and a similarity matrix was 

 constructed by using the Bray-Curtis similarity coef- 

 ficient (Clarke, 1993). Stress values are given for all 

 ordination plots; these values describe the quality of 

 the representation of multidimensional relationships 

 of the data in a two-dimensional plane. Stress factors 

 of less than 0.2 (<0.2 is considered to give a good rep- 

 resentation of sample "relatedness" and to prevent the 

 prospect of drawing false inferences) were obtained for 

 each ordination (Clarke. 1993). 



Analysis of similarities (ANOSIM) was used to test 

 whether fish assemblages in a priori groups differed sta- 

 tistically (Clarke, 1993). Abundance data for each spe- 

 cies were pooled for the four rockpools at each site and 

 time. Each ANOSIM comparison involved generating 

 4999 random permutations of the data to calculate the 

 probability that observed differences in the structure of 

 the fish assemblages among a priori groups could arise 



