202 



Fishery Bulletin 91(2). 1993 



APOGONIDAE 



GOBIIDAE 



17 ps 



17 PS 



LUTJANIDAE 



POMACENTRIDAE 



S 'andi 



3rd 



17 ' ps **y 



17 PS 



GOBIIDAE 



POMACENTRIDAE 



c. 



( ^»<S *** 



APOGONIDAE 



LUTJANIDAE 



_l 1 5 9 13 17 1 5 9 13 17 



Standard Length (mm) 



Figure 4 



Analysis of size structure in selected families of fishes collected by six sampling 

 methods on 3, 5, and 6 December 1986 off Lizard Island, Great Barrier Reef. (A) L„ 

 mean density/1000 m 3 (±SE) of four taxa in each of ten 2 mm size-classes collected 

 by purse seine (PS), bongo net (B), neuston net (N), and Tucker trawl (Ti. (B) L„ 

 mean density per sample (±SE) of gobiids and pomacentrids collected by light-trap 

 (LT) and light-seine (LSI. Size-classes as in (Al. (C) L n mean density per sample 

 (±SE) of apogonids and lutjanids collected by light-seine. Size-classes as in (Al. 



(Table 4). Only three significant 

 (p<0.05) differences were detected by 

 <-tests. The light-trap caught greater 

 numbers of large pomacentrids, the 

 light-seine greater numbers of large 

 atherinids and small gobiids. 



Among-night variation 



Larval and pelagic juvenile fishes 

 may vary in density at a particular 

 location over short time-periods rang- 

 ing from hours to days. We examined 

 the among-night variation in two 

 contexts. First, we used factorial 

 analysis to examine the variation at- 

 tributable to method of sampling and 

 sampling period (nights) in the net 

 collections. Second, we examined 

 the ability of nets and aggregation 

 devices to detect trends in density 

 of large individuals of some fami- 

 lies over a longer time-period (five 

 nights). 



A multivariate factorial analysis of 

 variance was used to examine trends 

 in mean density in six families: 

 apogonids, atherinids, gobiids, leth- 

 rinids, mullids, and pomacentrids. 

 Although both factors were signifi- 

 cant (Table 5), the significant interac- 

 tion between methods and nights 

 (Pillai's Trace F=1.65; df 36, 186; 

 p<0.01) indicates that differences 

 among methods were not consistent 

 over nights. 



Canonical Discriminant Analysis 

 was used to display the relationship 

 between methods and night of sam- 

 pling. Canonical variates 1 and 2 ex- 

 plained 93% of the variation in the 

 data set (Table 6). Figure 6 illustrates 

 the main conclusions from this analy- 

 sis. Tucker trawls, and neuston and 

 bongo nets each sampled a distinct 

 fish fauna with little among-night 

 variation. Purse-seine samples over- 

 lapped with those of the bongo nets 

 on two nights and were the most vari- 

 able, both within and among nights, 

 probably reflecting the influence of 

 few samples of small volume. Tucker 

 trawl samples were characterized by 

 consistently low numbers of the 



