206 



Fishery Bulletin 91(2). 1993 



undersampled smaller individuals, but was no better 

 than the bongo net at capturing larger larvae and pe- 

 lagic juveniles. This is consistent with the results of 

 Kendall et al. (1987) and Clarke (1991), who compared 

 bongo nets and larger trawls. The light-seine captured 

 a wide size-range of fishes because it combined the 

 sampling characteristics of both a purse-seine and an 

 aggregation device. 



Mesh size is an important determinant of catch com- 

 position because extrusion varies with mesh size. For 

 a given mesh size, extrusion is a function of body shape 

 and pressure across the net mesh (Clarke 1983 and 

 1991, Gartner et al. 1989). Body shape is species-spe- 

 cific, which emphasizes the importance of taxon-spe- 

 cific factors in methodological studies. Our results cover 

 a comprehensive range of body shapes, from slender 

 (gobiids) to deep bodied (apogonids and pomacentrids) 

 to moderately deep with elongate fin spines (lutjanids), 

 and should have general application. Purse-seines ap- 

 pear to herd planktonic organisms, while towed nets 

 actively filter, often under considerable pressure; thus 

 extrusion will vary between these two gear types re- 

 gardless of mesh size. As our primary interest was in 

 comparing a series of sampling devices in their normal 

 working configuration, we did not attempt to test the 

 effects of different mesh sizes within gear types. 



Although vertical stratification is minimal at night 

 in the study area (Leis 1986, 1991a), vertical distribu- 

 tion of the fishes could have affected apparent perfor- 

 mance of the samplers because each method sampled 

 somewhat differently in the vertical plane. Towed nets 

 were deployed at fixed depths. Experience elsewhere 

 has suggested that light-traps draw their catch from a 

 relatively narrow depth stratum, the upper 5m (P.J. 

 Doherty, unpubl.). However, only in the neuston net 

 can we confidently attribute greater catches (especially 

 of atherinids) to vertical stratification. For this study, 

 we assumed that vertical distribution of the fishes did 

 not affect our evaluation of the other methods. 



Horizontal or temporal variations in density may 

 also have confounded comparisons. A position effect 

 was possible because the aggregation devices were op- 

 erated at fixed positions about 700 m apart (Fig.l). A 

 temporal effect is possible because the bongo net and 

 Tucker trawl tows were run in blocks and not random- 

 ized during each night's sampling, although the order 

 of blocks was alternated among nights. 



Absolute sampling efficiency of the nets was not mea- 

 sured. Our estimates of sampling performance were 

 relative, because we did not obtain unbiased estimates 

 of the true densities of small pelagic fishes. We did not 

 attempt to use the methods of Somerton & Kobayashi 

 (1989) to correct our net catches because we felt some 

 of the assumptions required, especially those relating 

 to patch size and consistency through time, were not 



appropriate in the case of our study. The smaller bongo 

 net seemed to have equal or greater sampling effi- 

 ciency than the larger Tucker trawl at night for large 

 pomacentrids. 



A comprehensive comparison of the six sampling 

 methods would require two things. First, we would 

 need to standardize all results as number of organ- 

 isms per unit volume of water sampled. Second, we 

 would require an estimate of the sampling precision of 

 each device. For towed nets, both could be obtained 

 because flowmeters provided estimates of the volume 

 filtered for each tow. In the case of the purse-seine, it 

 was not possible to obtain reliable estimates of the 

 volume of water filtered during each deployment of 

 the net. Minor variations in the deployment procedure 

 can modify the dimensions of the volume enclosed by 

 the net. At present, we have no reliable way of esti- 

 mating this; therefore, for the purse-seine we have a 

 general estimate of water filtered based on idealized 

 dimensions of the deployed net. 



Volumes sampled by aggregation devices cannot be 

 estimated at this time, but preliminary calculations 

 (below) suggest they may be large. The bongo net as 

 operated in this study will sample -4000 nvVh, the 

 Tucker trawl -14,000 m 3/ h, and we estimate the light- 

 aggregation techniques could sample tens of thousands 

 of m3/h. Therefore, light-aggregation techniques may 

 be the best way to capture sufficient numbers of rarer, 

 larger stages for useful analyses. Aggregation meth- 

 ods may offer considerable advantages in studies of 

 settlement-stage reef fishes, but one must accommo- 

 date the characteristic taxonomic selectivity and un- 

 known sample volume. 



Two alternatives may explain the apparent dispar- 

 ity in numbers of larger pomacentrids estimated by 

 the bongo net (average 6.9/1000 m 3 ; Tucker trawl 

 catches averaged 1.49/1000 m 3 ) and the light-trap 

 ( average 273/h ): ( 1 ) The bongo net undersamples these 

 larger pelagic stages relative to the light-trap, or 

 (2) the light-trap samples larger volumes of water. As- 

 suming the two methods sample large pomacentrids 

 with equal efficiency, the light-traps sample volumes 

 on the order of 40,000m"h. This requires the trap to 

 capture, with efficiency equal to that of the net, pho- 

 topositive stages within a 7-50 m radius (to 5m depth) 

 of the trap, depending on the current speed (average 

 in the area is 15cm/s; Leis 1986) and geometry of the 

 light field. It is not possible to choose between alterna- 

 tives without a better measure of the effective volume 

 swept by traps. Work in progress will help resolve this 

 question. 



Short-term temporal variation in the density of par- 

 ticular families was more obvious in the results of some 

 methods than others. For the smaller size-classes, neu- 

 ston, bongo, and Tucker nets gave consistent results 



