LARSON and DeMARTINI: DISTRIBUTION OF FISHES IN KELP FOREST 



their more extensive bathymetric ranges, the low 

 biomass- density upper strata still contributed nearly 

 one-half of total biomass. 



The most abundant species at SOK were the cos- 

 mopolites (Tables 9, 10, 1 1). Sehorita, kelp bass, and 

 white seaperch comprised 82 and 90% of all 

 individuals in the kelp forests at SOK-D and SOK-U, 

 respectively. These species also contributed strongly 

 to overall integrated biomass, although large species 

 like California sheephead, barred sand bass, and 

 halfmoon were also important. As a result, the dis- 

 tribution of biomass among species was more even 

 than the distribution of numbers (Tables 9, 10, 11). 



Two relatively large fishes were more abundant at 

 SOK-D than SOK-U during fall of 1 979, contributing 

 to the differences (see below) in our estimates of total 

 biomass at each site (Table 1 1). The integrated abun- 

 dance of kelp bass was significantly higher, or nearly 

 so, at SOK-D (Numbers: t = 3.37, df = 7, 0.0K P < 

 0.02; Biomass: t = 2.65, df = 4, 0.05 < P < 0.1). 

 California sheephead were also more abundant at 

 SOK-D, as tested with log-transformed bottom data 

 (Numbers: t = 4.81, df = 6, P < 0.01; Biomass: t = 

 3.35, df = 5, 0.02 < P < 0.05) and with integrated 

 abundances (Numbers: t = 6.03, df = 5, P < 0.01; 

 Biomass: t = 4.92, df = 4, P < 0.01). Halfmoon 

 seemed to be more abundant at SOK-D, but the dif- 

 ference was not significant (Numbers: t = 1.78, df = 

 3, P > 0.1; Biomass: t = 1.78, df = 3, P > 0.1). 



At the kelpless cobble site, most fish were bottom 

 species and cosmopolites (Tables 5, 11). While 

 barred sand bass, black perch, and California 

 sheephead were fairly abundant in this area, the 

 average abundances of other species were less than 

 in the kelp-bed areas. The integrated numerical 

 abundance of all fishes was significantly lower in the 

 kelpless cobble area (cobble vs. SOK-U: t = 5.71, df 

 = 4,P< 0.01; cobble vs. SOK-D: t = 9.42, df = 3,P < 

 0.01; SOK-U vs. SOK-D: f = 0.79, df= 7, P> 0.4). A 

 one-way ANOVA of log-transformed counts on the 

 bottom showed significant differences among the 

 three areas (F 2 12 = 9.42, P < 0.01), but an a priori 

 comparison of SOK-U and SOK-D versus the cobble 

 area was not significant (F x 12 = 1.207, P > 0.25). 

 Thus, the lower overall numerical abundance at the 

 kelpless cobble area was due largely to the presence 

 offish above the bottom at SOK. The integrated total 

 biomass of fish did not differ significantly among the 

 three areas (F 2>11 = 0.25, P > 0.75), even though the 

 point estimate of 2.4 kg/100 m 2 at the cobble area 

 was lower than both values at SOK. However, barred 

 sand bass made up over 70% of fish biomass in the 

 cobble area, so most other species were much less 

 abundant there. 



We estimated the density of Macrocystis plants >1 

 m tall to be 7.51 ± 0.71 (1 SE) plants/ 100 m 2 at the 

 "kelpless" cobble area, 23.11 ± 1.47 plants/100 m 2 at 

 SOK-U, and 30.18 ± 1.69 plants/100 m 2 at SOK-D. 

 Thus, some kelp was present at the cobble area, but 

 the density of subadult-adult plants there was 25- 

 32% of density in our kelp-bed areas. 



DISCUSSION 



Sampling 



Regardless of water clarity, our camera and film 

 were unable to resolve fish beyond 3-4 m; this set an 

 upper limit of just over 1,000 m 3 to cinetransect 

 volume. Alevizon and Brooks (1975) noted that in 

 very clear, shallow waters, fish seemed difficult to 

 distinguish on film beyond 5 m. Ebeling et al. ( 1 980b) 

 found camera range to be 3-3.5 m at horizontal 

 visibilities of 4 and 15 m, and concluded that there 

 was essentially no relation between camera range 

 and horizontal visibility. Our data show this to be true 

 at visibilities >7-9 m. The fixed focal length of the 

 camera, shallow depth of field at maximum aperture, 

 and quality of film account for the limited camera 

 range, as discussed by Ebeling et al. (1980b). 

 However, our data show that camera range decreases 

 when visibility decreases to values that approach 

 maximum camera range. Corrections for visibility are 

 common in terrestrial line transects, whether the 

 area of a given transect is taken as fixed throughout or 

 as variable (Caughley 1977; Burnham et al. 1980). 

 We regarded the volume of a given cinetransect to be 

 fixed, its width determined by visibility. 



The relatively low upper limit to camera range may 

 help to make cinetransects in the water column more 

 accurate than visual censuses. Searching efficiency 

 would likely be poorer for broad visual transects 

 made to the limits of visibility. Furthermore, it is dif- 

 ficult to judge arbitrary smaller distances in open 

 water, unless they are only a meter or two on either 

 side of the diver. Cinetransects provide an almost 

 automatic upper limit to transect width, and this limit 

 is wide enough (about 3 m to either side in mod- 

 erately clear water) that a substantial volume of 

 water is censused. 



We have not verified the exact volume sampled in 

 each of our cinetransects, nor are we able to compare 

 densities measured in cinetransects with actual den- 

 sities (Brock 1982), since the latter have not been 

 measured by any method. To our knowledge, only 

 Keast and Harker (1977) have actually marked the 

 outside boundaries of visual underwater transects. 

 However, Terry and Stephens (1976) and Stephens 



49 



