Revisiting a comparison of larval reef fish composition in Opunohu Bay, Moorea, using sound and light traps

Caroline Engel Molly Zacher December 1, 2002 Revisiting a comparison of larval reef fish composition in Opunohu Bay, Moorea, using sound and light traps Abstract The abundance and species composition of larval reef fish were investigated in Opunohu Bay on the island of Moorea, French Polynesia. Light traps were used to catch larval samples every other night from November 12 to December 1, 2002 between 20:00 and 24:00 at four sites in and near Opunohu Bay. Comparisons between sites were made, with significant differences found between sites. Two of the sites were also sampled in 2000, and the results were comparable despite some variation in larval compositions. Amplified reef noises were projected in traps with and without light at two of the sites to test whether sound acts as an attractor for larvae coming into the reef at night. Sound was found to have no effect on larval catch. In addition, coral heads of Pocillopora spp. were made available for settlement to determine whether there was a correlation between larval catches in light/sound traps and arrival of larval damselfish on coral. Due to infrequent catches of the two Dascyllus species which settled on the corals, a correlation could not be found. Differences in composition and abundance between the four sites could be due to variation in habitat structure, local water conditions, and amount of light present. Introduction Most coral reef fish have a pelagic larval phase ending with settlement onto the reef. At settlement, larvae are conditioned by several factors including sensory cues and recruitment limitation/density dependence. In this study, we focused on these two questions using two different approaches. 1. Settlement Cues. Pelagic larvae are dependent on hydrodynamic transport; however, evidence suggests that many species also exhibit some form of active behavior, following sensory cues for settlement onto the reef (Leis & Carson-Ewart 1998). A subset of settling larval fish is attracted to light, so light traps are used to compare relative compositions and abundances of positively phototropic larvae over space and time (Bouley & Kaniewska 2000, Dufour & Galzin 1993). Light trap catches may be affected by variations in currents, the lunar cycle, turbidity, and water visibility. Because larvae have well developed hearing, sound may be an important settlement cue, and sounds within coral reefs are thought to be a beacon for larvae in presettlement

orientation to the reef (Fuiman & Myrberg 2002). This depends on whether larvae are capable of hearing sounds and whether the sounds possess sufficient energy above the lower threshold of hearing (Leis & McCormick 2002). In a study off the coast of New Zealand, significantly higher numbers of triplefin larvae were caught in light traps adjacent to underwater speakers projecting reef sounds than in light traps with no sound projection (Leis & McCormick 2002). Two years ago, UCSC students Paola Bouley and Paulina Kaniewska sampled two sites in Opunohu Bay using a light trap and detected significant differences in larval reef fish composition over time and between sites. A focus of this research was to test whether the patterns found by Bouley and Kaniewska were reproducible two years later (Caselle & Warner 1996). Moreover, this survey added two sites, covering a greater range of habitats, and investigated whether sound cues play a role in attracting larvae. 2. Recruitment Limitation/Density Dependence. A recurring question in coral reef fish compositional studies is whether observed population densities are a result of recruitment limitation or density dependence. The recruitment limitation hypothesis is a density independent mechanism that suggests that resident fish abundances rely solely on availability of larvae (Holbrook and Schmitt 1999). Density dependence affects recruitment rates when increasing densities of resident fish regulate subsequent establishment of new recruits (Hixon & Webster 2002). A focus of our study was to compare larval input, measured via light trap, to recruitment rates and density of recruits onto coral heads over time at one of our sites. If the light trap catches could be correlated with daily abundance of settled larvae on the cleared coral heads, light traps could then be used to predict larval input into the reef. Furthermore, density dependence could be inferred if populations on coral heads no longer receive new recruits despite continued larval input into the lagoon. Together, these methods could allow us to determine whether abundance of species settling on the coral heads was recruitment limited or density dependent. A number of current studies in marine ecology focus on what mechanisms regulate fish populations, however many provide only partial and/or theoretical answers. With the global collapse of fisheries and increasing extinction rates among marine species, understanding fish populations is crucial for conservation of marine species.

Materials and Methods Site Descriptions Moorea (17 30 S, 149 50 W) is a high volcanic island encircled by a barrier reef in French Polynesia. The barrier reef encloses a lagoon 800 to 1300 m wide, and the reef is intersected by several passes and two deep bays, Opunohu Bay and Cooks Bay, on the north coast. Breaking waves over the reef crest create a unidirectional flow of water into the lagoon, and this water returns to the ocean through the passes (Dufour et al. 1996). Input of pelagic larvae into the lagoon occurs via hydrodynamic transport across the reef crest. Figure 1. Map of Opunohu Bay with sampling sites BRS, LRS, WOS and CHS. Four sites within Opunohu Bay (see Figure 1) were selected for our research. The bay reef site (BRS, 17 30.342 S, 149 51.426 W) was located within West Opunohu

Bay at a green reef marker. This site lay directly over a reef with medium-high live coral coverage and is approximately 6 meters deep and next to a deep channel. BRS was subject to terrestrial runoff following storms and less water renewal than other parts of the lagoon. The lagoon reef site (LRS, 17 29.532 S, 149 51.095 W) was located in East Opunohu Bay at a black and white channel marker slightly east of White House Wall. Adjacent to a substantial reef, this site had moderate live coral cover with sand and dropped off steeply into a deeper channel than BRS. Our third site, the West Opunohu site (WOS, 17 29.311 S, 149 52.200 W), was closer to the crest in west Opunohu Bay and experienced the most water renewal. It was shallower than the other sites (approximately 4 meters deep) and was characterized by medium coral coverage and sand. Our fourth site, the Channel Site (CHS is located at 17 29.407 S, 149 52.303 W), was in deep channel water (approximately 16 meters deep) across from WOS. Light/Sound Trap Design Fish larvae were sampled at the four sites on Moorea using a light trap similar to the trap used by Bouley and Kaniewska. The light trap design consisted of a buoyant square (50 cm by 50 cm) of PVC pipe filled with hard foam held at the water s surface a clear waterproof plastic cylinder (see Figure 2). Inside the cylinder was a batteryoperated Coleman lantern, with two Princeton Tech dive lights zip-tied to the outside. Suspended below the light by ropes, a funnel-shaped plankton net (opening circumference 3 m, 1.7 meters deep) with a screw-top cylindrical cod end (12 cm diameter, 45 cm long) collected larvae as the light trap was hauled out of the water. Figure 2. Light trap design. A major sound cue for pelagic larvae to find the reef may be crashing waves, while a settlement sound cue may be coral reef sounds. At WOS, a mini-disc player (model Sharp 722) attached to a hydrophone was used to record sounds on a coral head (one that had many recruits, considered desirable for settlement), sounds of waves crashing outside the crest nearby, and ambient sounds near the water s surface. To make our sound trap, an underwater speaker was added to the light trap apparatus. The speaker, plugged into an amplifier, projected the recorded sounds at the water s surface above the trap s net. Kept in a plastic box to stay dry, the amplifier was powered by a

12-volt car battery connected to an inverter providing the necessary output of 110 volts. The 12-volt car battery was removed each night from Yannick and James blue Peugeot 205 car. Methods 1. Light and sound trap sampling Sampling occurred every other night of one lunar cycle between 8:00 pm and 12:00 am. At each of the four sites, the light trap was hung off the side of the boat twice for 10 minutes with a 5-minute break between sampling intervals. [A control using the trap without light would indicate which species of fish are present but not necessarily attracted to the light of the trap. One would then subtract from the light trap catches those fish that were caught in the control. However, we did not use this control because we wanted to look at overall fish caught, and did not plan to subtract out fish caught in the control. Also, we performed this type of control several times without catching anything, so the purpose of the control seemed moot.] At BRS, before using the light trap, sampling was done with sound traps, 10 minutes each for two different sound cues: coral head sounds, then ambient reef noises. Next the two light trap sampling intervals were done, followed by intervals using the two sounds in combination with light, giving a total of six sampling intervals. At CHS, the same sampling pattern was done as at BRS, but without ambient sounds, leaving a total of four sampling intervals. There were eleven sampling days using light traps, and three sampling days using sound. Catches from each treatment were placed in labeled Ziploc bags and brought back to the wet lab, where they were identified and counted. For some larval species, rearing was required to facilitate identification following metamorphosis. Salinity, turbidity, temperature, presence/absence of lunar light and currents were estimated at each site. 2. Recruitment to Pocillopora spp. Four Pocillopora spp. coral heads were cleared of fish and arranged (with fellow students Moira Decima and Holly Kindsvater) in the West Opunohu site, and the new recruits identified and counted each day. Recruits on these corals were not removed, so populations were allowed to accumulate over time. At least four other coral heads were set up and recruits counted and cleared every other day by Decima and Kindsvater for their experiment.

Using ANOVA in SYSTAT version 9, larval compositions from the four sampling sites were compared over the lunar cycle. Catches in the light traps at LRS and BRS were compared with data from 2000 to determine a lunar phase/site/year effect. The light trap and sound trap catches were compared to determine whether different fish compositions are drawn to light, light and sound, or sound only. Results 1. Site and day effects at four sites A multivariate analysis was performed to determine a day, site, and day-by-site effect among species over our four sites for 2002 (Table 1, Analysis 1). Nearly all taxa were included in the analysis, and the univariate F test showed significance (Table 2, Analysis 1) for all but five: Apogonidae, Albulidae, Microdesmidae, Hemiramphidae, and Soleidae. A stepwise discriminant analysis considering all taxa showed, based on larval compositions, LRS could be predicted 86% of the time, BRS 36%, CHS 32%, and WOS 41%. BRS, CHS, and WOS are most often mistaken for LRS. To detect broader patterns in species richness and diversity of larvae caught over time, a bar graph of mean number of species and mean number of individual larvae caught versus the lunar cycle was constructed (Figure 3). A scatterplot of number of species versus log 10 of individuals caught illustrates a linear relationship (Figure 4). 2. Comparisons to 2000 results To compare day-by-site effects with those of Bouley and Kaniewska, a day-bysite ANOVA was performed for LRS and BRS using the same 6 taxa as in 2000 (Table 2, Analysis 5). All 6 taxa were found to be significant in 2000; 4 of the 6 taxa were found to be significant in 2002 (Table 4). A discriminant analysis followed, exploring predictive ability for location based on taxa composition. Using species considered in 2000, BRS could be predicted with 100% accuracy, while LRS could be predicted only 41% of the time (Analysis 6). A stepwise discriminant analysis using all taxa from 2002 gave predictive abilities of 100% for BRS and 68% for LRS (Analysis 7). 3. Sound cues Using ANOVA, a comparison of treatments investigated whether a significant difference existed between light used alone and sound and light cues combined, and between different sound cues used with light (Table 1, Analyses 3 and 4). For both

number of species and number of larvae caught in the traps, there was no difference between light versus sound and light, nor between the two sound cues used in combination with light. Very few to no fish were caught with sound alone. 4. Recruitment to Pocillopora spp. The two larval species found settling on the Pocillopora spp. corals were Dascyllus aruanus and Dascyllus flavicaudus, consistent with observations by Holbrook and Schmitt. However, these species were infrequently caught in the light trap. Thus, no statistical analysis was performed to compare light trap catches to settlement on Pocillopora spp. coral heads. A monotonically increasing line would best approximate a graph of the abundance of these two species on the corals over time (Figure 5). Discussion 1. Site and day effects at four sites Larval abundance of 10 taxonomic groups was largely significant in the day-bysite multivariate analysis. An examination of bar graphs for these 10 groups revealed a strong pattern of site and/or day specificity for each. The 5 groups that were not significant lacked this relationship, and may have been affected more by other factors, such as current, turbidity, or salinity. Based on the discriminant analysis of larval composition, LRS could be distinguished from BRS, CHS, and WOS 86% of the time, and the other sites were confused most often with LRS. The reef next to LRS encompassed a wide range of habitats for coral reef fishes, being adjacent to a deep channel as well as a large reef. Thus, it follows that prediction of locality for the other sites could overlap with LRS. Hydrodynamics of the lagoon may also have played a role in the overlap of species composition of other sites with LRS. Further understanding of currents and water flux into Opunohu Bay and the lagoon would be useful. Also, it would be interesting to sample outside and inside the crest to see if there were differences in larval compositions across the crest. In looking at the four sites over the lunar cycle, some overall trends were apparent. Figure 3 shows that species richness and overall larval abundance exhibited two main peaks during the lunar cycle: a narrow one in the first quarter, and broader one in the third quarter moon. This is consistent with the larval input/lunar cycle patterns found in 2000. Another general relationship was identified in Figure 4. On average, for

every log 10 individuals, another species was found in the catch. Thus, one unit of species richness is added for every 10x units of larvae caught. It could be interesting to see if the same pattern holds true in adult coral reef fish populations. 2. Comparisons to 2000 results Larval abundances of 4 of 6 taxa used in the ANOVA showed a strong day-bysite effect in both 2000 and 2002. These were Chromis viridis, Dascyllus aruanus, Stegastes spp., and Chaetodontidae, all of which exhibited distinct lunar cycle peaks in the light trap catches. The D. aruanus peak occurred at the first quarter moon, while the others occurred near the third quarter moon. While influx of some species may simply vary from year to year, differences in trap construction/attraction and in weather conditions or current could account for some of the variability between years for the two species that lacked significance in the ANOVA for 2002. The trap used in 2000 was white and reflected light towards the surface better than did the green trap used in 2002. Also, low visibility may have inhibited the allure of the light trap on some sampling days. In the field, rain was observed to decrease visibility due to increased turbidity, so differences in weather patterns relative to the lunar cycle may have affected assemblages of larval species collected in 2000 versus 2002. Freshwater species, such as Gobiidae, are likely to be affected by salinity, and therefore by rain (Bouley and Kaniewska 2000). Strong current swept less tenacious larvae away from the trap, even if they were present and attracted to the light. If a strong current were present at different phases of the moon for 2000 versus 2002, larval compositions in the light trap catches between years would have been affected. The species that were good predictors of locality in 2000 succeeded in predicting for BRS 100% of the time in 2002, but a different assemblage of species was required to obtain a >50% predictability for LRS. Again, the species compositions over time and sites may simply differ, or they may have been influenced by weather conditions, current, and sampling methods. 3. Sound cues Results indicated that projected sounds neither detracted nor added to the larval catches. A more complete survey of sounds at the sites and of intensities of projected sounds relative to natural sounds might have been useful. There may exist some sounds that act as attractants at one intensity but repel larvae or do nothing at another intensity.

Limitations in output frequency range and resolution of sounds on the underwater speaker could have affected the quality of the sound cues (Fuiman & Myrberg 2002). Clearly the sound cues, as tested here, were ineffectual. Numerous technical difficulties inhibited sampling with sound, limiting the total number of sound samples to three. An effect might have been detected had more sound sampling been performed. 4. Recruitment to Pocillopora spp. The monotonically increasing line graph of Dascyllus spp. accumulated on the corals versus time (Figure 5) indicated that the system had not yet reached density dependence, but rather remained in a recruitment limitation phase (Hixon & Webster 2002). It is likely that more than one lunar cycle was required for coral heads to become fully stocked. Also, the light trap catches did not seem to correlate with the input of Dascyllus spp. onto the corals. Almost no D. flavicaudus were caught despite influx of larvae, indicating a lack of attraction of that species to the light trap. D. aruanus did begin to settle on the corals during the first quarter moon phase, when they were being caught in the light trap. According to Moira Decima and Holly Kindsvater, around the third quarter moon there was another influx of D. aruanus on the corals that were cleared daily. This influx, however, was not detected in the light trap catches. Perhaps another method of catching larvae may be more applicable to studies involving D. aruanus and D. flavicaudus.

Number of species and Number of individuals vs. Lunar Cycle 90 60 Value 30 0 0 10 20 30 Lunar Cycle (0 = new moon) # species # individuals Figure 3. Species richness and overall larval abundance over one lunar cycle.

12 10 NSPECIES 8 6 4 2 0 20 40 60 80100 NIND Figure 4. Number of species versus log 10 number of individuals caught in light trap.

9 8 7 Sum of recruits 6 5 4 3 2 1 0 0 5 10 15 DAY Figure 5. Accumulation of D. aruanus and D. flavicaudus on corals over time. Courtesy Moira Decima. Table 1. Summarizes outputs for GLM and ANOVA analyses. Analysis Source Pillai df p value 1 Day 5.166 140, 400 0.000 1 Site 2.126 42, 99 0.000 1 day*site 7.750 420, 616 0.000 3 treatment n.a. n.a. 0.959 (ambient+light vs. light) 3 treatment n.a. n.a. 0.999 (coral+light vs. light) 3 treatment n.a. n.a. 0.936 (coral+light vs. ambient+light) 4 treatment n.a. n.a. 0.970 (ambient+light vs. light) 4 treatment n.a. n.a. 0.995 (coral+light vs. light) 4 treatment n.a. n.a. 0.926 (coral+light vs. ambient+light) 5 day*site (C. viridis) n.a. 10 0.000 5 day*site (D. aruanus) n.a. 10 0.000 5 day*site (Stegastes spp.) n.a. 10 0.001 5 day*site (Albulidae) n.a. 10 0.144 5 day*site (Chaetodontidae) n.a. 10 0.000 5 day*site (Gobiidae) n.a. 10 0.091 Table 2. Indicates which categories were included in GLM, ANOVA, and discriminant analyses. Analysis Category Included (+) 1 2 3 4 5 6 7 # individuals + # species +

Chromis viridis + 0.000 + + + + ACANTHURIDAE + 0.001 + + Acanth. triostegus + Zebras. scopas + APOGONIDAE + 0.132 + + Dascyllus aruanus + 0.000 + + + SCORPAENIDAE + 0.041 + + + GOBIIDAE + 0.001 + + + + CHAETODONTIDAE + 0.000 + + + Stegastes spp. + 0.000 + + + + SYNODONTIDAE + 0.003 + + ALBULIDAE + 0.103 + + + + MICRODESMIDAE + 0.060 + + HEMIRHAMPHIDAE + 0.608 + + SOLEIDAE + 0.492 + + LUTJANIDAE + 0.003 + + Fistularia commersoni + + Table 3. Describes GLM, ANOVA, and discriminant analyses (as referenced in Tables 1 and 2). Analysis 1 GLM for day, site, day*site effect for our four sites. Analysis 2 Discriminant analysis using all larval spp at all four sites. Analysis 3 ANOVA for # species for sound and light vs. light only. Analysis 4 ANOVA for # individuals for sound and light vs. light only. Analysis 5 ANOVA for day*site effect at LRS and BRS for comparison to 2000. Analysis 6 DA using taxa from 2000 list at LRS and BRS. Analysis 7 DA using all larval taxa at LRS and BRS. Table 4. Comparison of ANOVA univariate analysis for 6 taxa in 2002 and 2000 (Analysis 5). 2002: p value, df 2000: p value, df Chromis viridis 0.000, 10 0.002, 11 Dascyllus aruanus 0.000, 10 0.000, 11 Stegastes spp. 0.001, 10 0.041, 11 Albulidae 0.144, 10 0.050, 11 Chaetodontidae 0.000, 10 0.000, 11 Gobiidae 0.091, 10 0.014, 11 Acknowledgements The light trap was kindly donated to us by Gilles Lecaillon. Ron Schusterman, Michael Poole, and Dan Costa generously lent us sound equipment. Giacomo Bernardi, Pete Raimondi, Alain Lo-Yat, your instruction and advice proved invaluable, and your company was much appreciated. Moira Decima, Holly Kindsvater, Sara Worden, Pete Dal Ferro, Jonna Engel, Mark David Readdie, Dawn Jech, thanks for your help and for fun in the field. James Algier, Yannick Chancerelle, Hinano, and Prince Cookies, we

enjoyed getting to know you. A special thanks to Corey Phyllis and Damon Wolf for their cooperation on the boat. References Bergenius, M.A.J., Meekan, M.G., et al. (2002). Larval growth predicts the recruitment success of a coral reef fish. Springer-Verlag website. Bouley, P. & Kaniewska, P. (2000). Comparison of larval reef fish composition between a bay reef site and a lagoon reef site at Moorea, French Polynesia. UCSC internal publication. Caselle, J. & Warner, R. (1996). Variability in recruitment of coral reef fishes: the importance of habitat at two spatial scales. Ecology. 77, No. 8: 2488. Dufour, V. & Galzin, R. (1993). Colonization patterns of reef fish larvae to the lagoon at Moorea Island, French Polynesia. Marine Ecology Progress Series. 102: 143-152. Dufour, V., Lo-Yat, A. & Riclet, E. (1996). Colonization of reef fishes at Moorea Island, French Polynesia: Temporal & Spatial Variation of the Larval Flux. Marine Freshwater Res. 47: 413-22. Fuiman, L. A. & Myrberg, A. A. (2002). The sensory world of coral reef fishes. Coral Reef Fishes (P. F. Sale, ed.), pp. 123-148. Academic Press, San Diego, CA. Hixon, M. A. & Webster, M. S. (2002). Density dependence in reef fish populations. Coral Reef Fishes (P. F. Sale, ed.), pp. 171-199. Academic Press, San Diego, CA. Holbrook, S. J. & Schmitt, R. J. (1999). Settlement and recruitment of three damselfish species: larval delivery and competition for shelter space. Oecologia. 118:76-86. Leis, J.M. & Carson-Ewart, B.M. (1998). Complex behaviour by coral-reef fish larvae in open-water and near-reef pelagic environments. Environmental Biology of Fishes. 53: 259-266. Leis, J.M. & McCormick, M. I. (2002). The Biology, behavior & ecology of the pelagic, larval stage of coral reef fishes. Coral Reef Fishes (P. F. Sale, ed.), pp. 171-199. Academic Press, San Diego, CA. Lo-Yat, Alain 2002. Presentation 11/13/02, CRIOBE, French Polynesia. Colonization Patterns of Larvae in French Polynesia.