Choat et al.: Comparison of ichthyoplankton sampling methods 



207 



over short time-periods (Fig. 6). Catches from the purse- 

 seine were more variable within a sampling period 

 and showed greater variability among nights of sam- 

 pling than did the towed nets. This reflects the local- 

 ized sampling area and small sample volume of the 

 purse-seine. For larger mullids and pomacentrids, simi- 

 lar trends in density over five nights were identified 

 by the aggregation devices. These trends were not ap- 

 parent in the data from the towed nets. Thus, the 

 aggregation devices seem particularly suited to stud- 

 ies of short-term temporal variation in the larger 

 (>6mm) size-classes. The rapid and independent 

 changes in density of the larger individuals of these 

 two families suggest that larger pelagic stages are not 

 present in the water at all times at a location. The 

 alternative, that there are short-term taxon-specific 

 changes in catchability due to changes in behavior of 

 the fishes, seems less likely, but cannot be dismissed 

 without further study. 



A number of other studies have compared sampling 

 methods for planktonic and pelagic assemblages. Purse- 

 seines were found to be superior to towed nets for 

 sampling larval anchovies (Murphy & Clutter 1972). 

 Larger, faster, more-transparent nets may minimize 

 net avoidance (Clutter & Anraku 1968). However, Smith 

 & Richardson (1977) suggest that increased net size 

 and towing speed may intensify the disturbance in 

 front of the net and increase net avoidance. All towed 

 nets in these cited studies employed towing bridles, 

 which are a source of water disturbance and, thus, net 

 avoidance by fishes. Towing bridles were not used in 

 the present study, which may be why our conclusions 

 differ from those of Clutter & Anraku (1968) and 

 Murphy & Clutter ( 1972 ). 



We agree, however, with Clarke (1991) who made 

 detailed comparisons of the effectiveness of two types 

 of bongo nets and a midwater trawl in capturing reef- 

 fish larvae. He suggested that the bongo nets (0.7m 

 diameter with 0.183 mm mesh, and 1.25 m diameter 



with 2.5 mm mesh) sampled larvae as well or better 

 than a 3 m Issacs-Kidd trawl (6 mm mesh). Clarke con- 

 cluded that when densities of larvae were high, 0.7 m 

 and 1.25 m bongo nets were the most effective meth- 

 ods for sampling small and large larvae, respectively. 

 Although larger nets are assumed to capture more and 

 larger fishes due to lessened avoidance (Clarke 1983 

 and 1991, Methot 1988), this was not true in our study 

 nor is it always true in other pelagic groups (Barnes & 

 Tranter 1965, Sands 1978, Pillar 1984). 



One other significant study compared catches from 

 a light-trap with those from a towed net. Gregory & 

 Powles (1988) investigated a relatively simple plank- 

 tonic assemblage of freshwater fishes. Based on a com- 

 parison of taxonomic composition and size of fishes, 

 they concluded that both sampling methods should be 

 used to avoid selectivity biases. An interesting conclu- 

 sion that differs from our results was that the light- 

 trap provided a better representation of size-classes, 

 including smaller individuals, than did the towed net. 

 This emphasizes the taxon-specific and, perhaps, habi- 

 tat-specific nature of gear-performance measures. 



We agree with Omori & Hamner (1982) that the 

 sampling device and program selected must be ques- 

 tion-driven (Kingsford 1988). In order to assist in the 

 choice of appropriate methods, we summarize the per- 

 formance and sampling properties of the six methods 

 employed in this study (Table 7). Surveys of larval 

 fishes are best accomplished with a bongo net. This 

 will cover a significant portion of the size-range in 

 many important taxa, including larger individuals, at 

 least at night. No extra benefits were apparent from 

 using the larger Tucker trawl. A major advantage of 

 bongo nets is the relative ease with which they may be 

 deployed and retrieved. As expected, neuston nets fo- 

 cused on neustonic fishes. 



Surprisingly, the purse-seine provided results com- 

 parable to the bongo net despite the small volumes 

 sampled. Among-sample variances were predictably 



