406 



Fishery Bulletin 100(3) 



Light traps 



We deployed three light traps at each site on each day, from 

 30 July to 11 August 1997. Each of the three traps at a site 

 was a different modification from two basic designs, with 

 each trap rotated among stations within a site to account for 

 any trap- or location-specific biases. One trap design had a 

 65-cm diameter polycarbonate top and 1.5-mm nylon mesh 

 enclosing a volume of 200 L, with four clear plastic funnels 

 as fish entrance holes and a fluorescent light powered by 

 nickel-cadmium rechargeable batteries (see Sponaugle and 

 Cowen, 1996b). A second trap design was similar but had 

 an 80-cm diameter wood top instead of a polycarbonate top, 

 a larger volume of 360 L, twice the number of entrance 

 holes, a fluorescent light powered by rechargeable sealed 

 lead-acid batteries, and an automatic timer for turning on 

 the light remotely. The third trap design was composed 

 of a rectangular plexiglas trap (42 cm x 38 cm x 70 cm) 

 with rigid plexiglas panels, a plastic tray, and four entrance 

 slots, constructed to enclose a volume of 110 L based on 

 the design of Munday et al. (1998). but having the same 

 electronics design as that for traps made with polycarbon- 

 ate tops. All light traps were placed 3^ m below the sea 

 surface, suspended from moorings with subsurface buoys, 

 and stationed approximately 12-15 m in front of each reef 

 and 50-60 m from each other in a linear array along the 

 offshore edge of the reef Traps were deployed between 1730 

 and 1830 h and retrieved the following morning between 

 0900-1030. All fishes (except for ubiquitous clupeids and 

 atherinids) were collected and placed in vials of 70*^ or 95% 

 ethanol, identified to family, and later measured with ver- 

 nier calipers to the nearest 0.1 mm standard length (SL). 



Channel nets 



We deployed one surface (0-1 m depth) and one subsurface 

 (2^ m depth) channel net at each site on each day (30 July 

 to 11 August 1997) at a distance of approximately 50 m 

 offshore of the center of the light-trap array. The channel 

 nets were based on the design of Shenker et al. (1993), 

 had a mesh size of 2 mm, and were positioned 30 m apart. 

 The nets were suspended from surface buoys moored to 

 concrete blocks or mooring anchors, allowing them to turn 

 and fish both ebb and flood mixed-semidiurnal tides. The 

 subsurface net mouth opening was 2 m wide x 2 m high. 

 The mouth of each surface net was 2 m wide x 1 m high 

 and was equipped with a General Oceanic Model 2030R2 

 flow meter and low-speed rotor blade suspended in the 

 mouth opening. Flow-meter readings were recorded from 

 the surface channel net to estimate relative current veloc- 

 ity between the two sites. All nets were equipped with PVC 

 (polyvinylchloride) rods along the length of the netting to 

 prevent entanglement during slack tides, and the codends 

 were constructed to sink and close the end of the net to 

 contain fish larvae during times of very low current veloc- 

 ity. Nets were deployed and retrieved at approximately the 

 same time of day as the light traps. Channel nets were not 

 sampled at dusk to distinguish catches during the day from 

 the following night because previous studies indicated that 

 daytime catches account for a very minor percentage of the 



total number of fish transported onto the Great Bahama 

 Bank (Shenker et al., 1993; Thorrold et al., 1994c). At the 

 laboratory, samples were rough-sorted to remove debris, 

 fixed with 10% formaldehyde for 24—72 h, and then trans- 

 ferred to 70% isopropyl alcohol for later identification. Fish 

 were measured with vernier calipers to 0.1 mm SL. 



Analysis 



To compare the number of taxa collected by Ught traps and 

 channel nets between sites and for both sites combined, we 

 summed the total number of families caught by each sampling 

 device to calculate family richness. We also used the Brillouin 

 index of species diversity (Magurran, 1988) to compare the 

 diversity of families of fish larvae between sampling methods 

 and sites. This index is preferable to the Shannon-Weiner 

 index because samples collected by light traps and channel 

 nets are nonrandom; for example, light traps produce biased 

 samples based on the sensitivity of species to a light cue. 



To compare the relative abundance of families caught 

 by light traps and channel nets between sites and for both 

 sites combined, we standardized catches in channel nets to 

 the number of larvae per 1000 m-^, using flow meter read- 

 ings recorded each day, and we standardized catches from 

 light traps as the number of larvae per day We could not 

 standarize catches to catch per unit of effort because the 

 length of time that the lights were operational was vari- 

 able and dependent on the type of light device (see heading 

 "Light traps") and variance in battery life. Moreover, the 

 volume over which light traps attract larvae is difficult to 

 quantify, especially when external factors such as current 

 velocity may largely affect catch rates (Thorrold, 1992; 

 Meekan et al., 2000), and Meekan et al. (2000) suggested 

 that it is useless to convert catch rates into densities. 



We used Spearman's rank correlation coefficient (Zar, 

 1984) to compare the relative abundance of taxa that repre- 

 sented at least 1% of the catch (Cheat et al., 1993) for either 

 sampling device. We also used this correlation coefficient to 

 compare the relative abundance of taxa caught by surface 

 and subsurface channel nets between sites for taxa that 

 represented at least 1% of the catch for either net. In order 

 to determine whether there were significant differences in 

 the mean, median, and maximum length of families offish 

 larvae between light traps and channel nets, we used a Wil- 

 coxon paired-sample test (Zar, 1984) in which differences in 

 length for each family caught by both sampling devices were 

 ranked. Finally, we used a t-test to test for significant differ- 

 ences in current velocity and the proportional abundance of 

 total larvae caught by each sampling device between sites. 



Results 



Richness, diversity, relative abundance, 

 and individual size 



According to our hypotheses in regard to active and pas- 

 sive collection of larvae, both family richness and diversity 

 would differ between light traps and channel nets. A total 



