EBELING ET AL.: ANNUAL VARIABILITY OF REEF FISH 



the canopy habitat, the effect of more variable 

 light levels is apparently not severe. A similar 

 regression analysis of pooled 1971-72 island- 

 canopy samples in = 60) was nonsignificant (P = 

 0.9), though visibility ranged from 4.6 to 10.7 m. 



But while such inferences from camera optics 

 and counts-visibility relations indicate that esti- 

 mates of relative abundance obtained from cine- 

 transects are comparable over a wide range of 

 sampling conditions, we feel that the calibration of 

 absolute densities presents difficulties. The pri- 

 mary problem is that, particularly in the kelp 

 canopy, our estimates of area or volume sampled 

 are tenuous. For this reason, we used absolute 

 densities only for computing Wolda's (1978) mea- 

 sure of annual variation, which required esti- 

 mated abundances per species, standardized for 

 differences in sampling effort among years and 

 between canopy and bottom habitats. We feel that 

 this is proper because a systematic error in es- 

 timating will have little effect on the measure's 

 value, which is based on yearly ratios of popula- 

 tion sizes, not on sizes per se. 



Also, combining canopy and bottom transects 

 may miss some fish at middepth. Canopy transects 

 covered depths between 1 and 4 m, which included 

 the greatest concentration of fish in the upper 

 water column. Bottom transects covered depths 

 between the reef and about 2 m upward, which 

 included greatest concentrations in the lower col- 

 umn. Nonetheless, over bottoms averaging 8.5 m 

 deep, cinetransects generally missed the top meter 

 as well as midwater between 4 and 6.5 m. Hence, 

 our fish counts, even of overt midwater species, 

 probably underestimated true abundances. 



Annual Variability 



Species composition (order of relative or ranked 

 species abundances), rather than richness 

 (number of species), contributed most to differ- 

 ences between mainland and island fish as- 

 semblages, which were most marked for the bot- 

 tom assemblages (see also Ebeling et al. in press). 

 Yearly mainland and island samples had the same 

 number of species, although island species diver- 

 sity was slightly greater because individuals were 

 more evenly distributed among species. At the 

 island site, species in a "kelp-rock" habitat group 

 (Ebeling et al. in press) — tropically derived 

 species such as Pimelometopon pulchrum and G. 

 nigricans — were relatively more abundant than 

 at Naples Reef. We had no indication that fishing 



intensity for such species (spear fishing, bottom 

 angling) differed between the two sites. Nor was 

 unnatural disturbance by kelp harvesting a fac- 

 tor. Furthermore, virtually unexploited kelp-rock 

 species, such as Hypsypops rubicundus , were rela- 

 tively more abundant at Santa Cruz Island. This 

 indicates that much of the mainland-island differ- 

 ence in species composition probably reflected the 

 observed differences in structure of natural 

 habitat rather than differences in exploitation. 

 Likewise, but on a broader scale, insular and con- 

 tinental shore-fish faunas are distinguished in the 

 tropical western North Atlantic (Robins 1971; 

 Gilbert 1972). Whereas turbid waters, muddy- 

 silty bottoms, and few reefs characterize mainland 

 habitats, clear water, coral reefs, and more stable 

 conditions typify island habitats. Consequently, 

 island fish assemblages contain relatively more 

 specialized reef species, such as pomacentrids and 

 labrids, that require the trappings and provisions 

 of complex surfaces. 



We felt that composition and abundance of the 

 fish assemblages remained fairly constant among 

 years, considering that they inhabit a presumed 

 zone of faunal transition (Hubbs 1974; Horn and 

 Allen 1978). Species composition varied more be- 

 tween sites and habitats than among years. This 

 indicates that a particular assemblage persists, 

 despite significant yearly variation in its fish and 

 species counts. Yet we had few standards for com- 

 parison. Sale (1978:85) knew of no evidence that 

 demonstrated "long-term local stability in reef 

 fish communities," presumably because long-term 

 monitoring studies were wanting. With a 7-yr 

 study of fishes inhabiting a rocky tidal pool in the 

 northern Gulf of California, however, Thomson 

 and Lehner (1976) showed that fish abundance, 

 diversity, and species order were seasonally pre- 

 dictable and varied little from year to year. In fact, 

 the fish assemblage was remaikably resilient, re- 

 covering quickly from unpredictable and devastat- 

 ing disturbances, such as severe storms, winter 

 kills, and rotenone poisoning. More general- 

 ly ,Wolda (1978) emphasized the need for actual 

 measures of annual variability in tropical and 

 other animal assemblages to test a plethora of 

 theoretical speculation. 



Our values of Wolda's ( 1978) measure of annual 

 variation (AV) in arrays of species were in fact 

 relatively low. AV's for fish assemblages at main- 

 land and island sites (0.11, 0.15) did not differ 

 significantly (F-tests of variance ratios) from most 

 of those (0.06-0.33, averaging 0.15) for arthropods 



373 



