LARSON and DeMARTINI: DISTRIBUTION OF FISHES IN KELP FOREST 



strata; W h , the weighting factor; and n h , the number of 

 days sampled in stratum h. The portion of the for- 

 mula included in the summation is the usual estimate 

 of variance for stratified means (Snedecor and 

 Cochran 1980), and the root of this sum is the stan- 

 dard error of mean density (per 1,000 m 3 ) throughout 

 the water column. Multiplying by (1,500/1,000) 2 

 adjusts the standard error for the larger volume of 

 water in the column over 100 m 2 . 



Estimates of integrated abundance at the kelp- 

 depauperate site were obtained by converting mean 

 density on the bottom to mean density over 100 

 m . 



Arithmetic means (of untransformed data) were 

 used for all estimates of density and abundance. 

 Geometric means (obtained by back-transforming 

 the means of log-transformed data) underestimate 

 absolute densities in a manner proportional to their 

 variances. Adjustments for this underestimation 

 (Elliott 1971) are usually based on the assumption of 

 log-normal distributions, and we could not make such 

 an assumption. However, some statistical com- 

 parisons were made with log-transformed data to 

 avoid the problem of heterogeneous variances. 

 These were comparisons of mean numbers and 

 biomass on the bottom, where varying transect 

 volume did not confound the calculation of variance. 

 Other comparisons, however, were made with 

 untransformed data. These included tests for dif- 



ferences in numbers or biomass in the above-bottom 

 strata and in the entire water column. When all three 

 areas were compared, a one-way ANO VA was used if 

 variances were not heterogeneous. T'-tests for un- 

 equal variances (Bailey 1959) were used for pairwise 

 comparisons of areas when variances were un- 

 equal. 



RESULTS 

 Cinetransect Calibration 



We estimated cinetransect length to be about 76 m. 

 Six down-current trials averaged 78.3 m in length 

 (standard error (SE) = 1.5 m, range = 74-82 m), 6 

 upcurrent trials averaged 72.8 m in length (SE = 2.3 

 m, range = 67-82 m), and the overall average was 

 75.6 m (SE = 1.5 m). 



Camera range was an asymptotic function of 

 horizontal visibility, with little increase in camera 

 range at visibilities beyond 7-9 m (Fig. 3). Camera 

 range was appreciably lower when the camera was 

 facing the sun than vice versa, particularly at greater 

 visibilities. This was reflected in each of the curves fit 

 (Table 3). Since divers did not record whether actual 

 transects faced into or away from the sun, we used the 

 curve fit to all camera range-horizontal visibility 

 values to calibrate cinetransect volume. The logistic 

 equation provided, by slight margin, the best fit to 



O 3 



< 



2 . 



< 



1 



284 + 1 893(0.582 * 



<8> 



* 



-*e e- 



• INTO SUN 



: AWAY FROM SUN 



1 2 3 4 5 6 7 8 9 10 11 12 13 



HORIZONTAL VISIBILITY (m) 



Figure 3.— Relation of camera range (the distance at which fish could be distinguished on film) 

 and horizontal visibility. Points are observations of maximum camera range at different visi- 

 bilities with the camera facing into and away from the sun. The equation and line show the logistic 

 function fit to these points. 



43 



