nSHERY BULLETIN: VOL. 78, NO. 2 



between years than between habitats or sites, 

 although canopy assemblages maintained less 

 site-specific integrity than bottom assemblages. 

 Coincident peak abundances of several species at 

 both sites in 1972 contributed to significant yearly 

 differences in fish counts, although significant 

 year-site interactions revealed exceptions to the 

 generally concordant annual trends. Fishes that 

 aggregated in the canopy habitat, especially mid- 

 water planktivores, probably contributed most to 

 annual variation measured by between-year log 

 ratios of numbers per species. Yearly differences 

 in fish abundances were loosely related to water 

 clarity and temperature and kelp density, but 

 correlations were not clear-cut. 



DISCUSSION 



Sampling 



With limited time, personnel, and budget, visu- 

 al transecting may be the most appropriate 

 method for sampling fish populations in the com- 

 plex reef environment, so long as it is understood 

 that this method always underestimates densities 

 of small, hidden, and/or cryptic species (Brock 

 1954; Jones and Chase 1975). Although some in- 

 vestigators aver that destructive methods ( poison- 

 ing, dynamiting) provide broader sampling (Ran- 

 dall 1963; Goldman and Talbot 1976), others 

 counter that visual methods are more representa- 

 tive because they record individuals of larger, 

 stronger species that escape the slaughter (Smith 

 and Tyler 1973). Hence, a thorough census of cov- 

 ert and overt species probably requires both 

 methods (Quast 1968c). 



Cinetransecting is analogous to visual transect- 

 ing. Both methods may miss most covert fish 

 ( Alevizon and Brooks 1975), but record most overt 

 individuals. For example, the rank order of species 

 abundances from all mainland-bottom samples 

 (Table 1) correlated significantly (tau = 0.66, 

 P<0.001) with that of daytime visual transects 

 made along a transect line about the reef crest at 

 this site throughout the year (Ebeling and Bray 

 1976: table 3). Four of the five top-ranking species 

 were the same in both studies, even though cine- 

 transects covered a much broader area. 



However, cinetransects have some advantages 

 over visual transects. They can be made quickly, 

 as many as 50/ d in the present study. Cinetran- 

 sects provide permanent records of the fish and 

 their environment, not only for greater accuracy 



in identifying and counting fish, but also for reuse 

 in related studies (see Alevizon 1975; Bray and 

 Ebeling 1975; Love and Ebeling 1978 1. Diver pho- 

 tographers can proceed slowly and steadily, not 

 diverting their attention from sampling to record 

 observations or follow a transect line. They do not 

 need extensive training in quick recognition of 

 fish species and numbers, so can be replaced by 

 others if required; if cinetransect samples are 

 sorted into subsets, each filmed by one or the other 

 of two different divers, correlations between the 

 corresponding diver-specific species arrays are 

 very high. For example, the four habitat-site- 

 specific samples filmed by two divers in 1973, 

 when sorted to eight diver-specific subsets, gave 

 tau rank correlations ranging from 0.71 to 0.'89 

 (P<0.001). 



Within broad limits, furthermore, water visibil- 

 ity and light levels probably do not appreciably 

 affect the volume of water sampled by cinetran- 

 sects filmed along the bottom. At a given focus 

 distance, the camera lens' depth of field is in- 

 versely related to the diameter of its aperture. In 

 bright light, the aperature is small, creating a great- 

 er zone in which objects are in focus. In the kelp 

 forest, however, light was generally so dim, even 

 on clear days, that the aperture was almost always 

 fully open. Thus, shading probably creates a fairly 

 constant depth of field. To check this, we estimated 

 the distance at which objects were first identifiable 

 on film. Two divers swam along a tape measure 

 ending in a fishlike target, one filming the target 

 and nearby fish, the other signaling distance from 

 target. During the first trial when underwater 

 visibility (distance at which target was discern- 

 ible) was 15.2 m, fish were identifiable on film only 

 when photographed within about 3.5 m of the 

 camera. During the second when visibility was 

 only about 4.0 m, fish were still identifiable when 

 photographed within about 3 m. Hence, the fairly 

 constant depth of focus of the camera's lens, which 

 was always set at 2.0 m on the distance scale, 

 standardized the maximum distance ( about 3.5 m) 

 at which a photographed fish was identifiable. 



Linear regressions of logio-transformed fish 

 counts on estimated underwater visibility provide 

 further evidence that visibility had little effect on 

 values. Mean fish and species counts for the 1973 

 and 1974 island-bottom samples were similar, so 

 the two were combined as one large sample in = 

 103) for regression analysis. Although visibility 

 varied between 2.1 and 15.2 m, the regression was 

 nonsignificant ( ANOVA F-test, P = 0.3). Even in 



372 



