544 



Fishery Bulletin 104(4) 



Longitude ("W) 



t20"30' 120'20' 120"10' 12000' 119 50' 119 40' 119 30' 119 20' ll19"10' 



I Irene ^^ 

 i 73m "W 



 Hidalgo -tf 



: 130m ^^ 





Figure 1 



Location of platforms in the Santa Barbara Channel and Santa Maria Basin. Platforms discussed in this 

 paper are shown as black stars and their bottom depths are indicated. The open circles indicate other 

 platforms. 



65) and 50,048 fish of the original 52,999 observations, 

 or 95.1% offish observed on all dives. In our analyses, 

 we separated bocaccio (Sebastes paucispinis) and ling- 

 cod into two categories, YOY and older fishes, based on 

 length-at-age and length-at-first-maturity data (Miller 

 and Geibel, 1973; Cass et al., 1990; Love et al., 2002). 

 We used chi-square goodness-of-fit tests to determine 

 which species tended to avoid or favor certain beam 

 habitats. Species tend to be associated with particular 

 depth ranges, platforms are placed at fixed depths, and 

 some species may or may not have ever been observed 

 on some platforms. Thus, the proportion of a particular 

 beam habitat available to a given species was deter- 

 mined as the proportion of that habitat occurring only 

 on those platforms where the species was observed. We 

 hypothesized that if a particular species does not favor 

 or avoids certain habitats, the expected number of those 

 fish seen in particular habitats would be proportional 

 to the amount of available habitat. In equation form; 

 Let u^^ be the proportion of gap j ( j = 0, 1, 2, 3, or 4) 

 available to species / (; = 1, 2..., 27) and T, be the total 

 number individuals of species ;' observed. Then, the 

 expected number offish / at gapj, E^^ = r/ Y T,. Under 

 the null hypothesis for species (', 



■' (O 



E f 



J,' 



j=o Ej,, 



is distributed as a chi-square random variable with 

 n,-l degrees of freedom, where /;, is the number of 

 Uy>0. The asymptotic assumptions for the chi-square 

 test are not valid if the expected value of many cells 

 is small. Cochran (1954) developed a conservative rule 

 of thumb that the test not be used if more than 20% of 



the expected cell frequencies are less than five. Koehler 

 and Larntz (1980) suggested that the chi-square test is 

 reasonable if the total number of observations is greater 

 than 10, the number of categories is at least 3, and the 

 square of the number of observations is greater than 10 

 times the number of categories. 



If the goodness-of-fit hypothesis was rejected, we ex- 

 amined the individual deviations, X^, = (O^, - £ )/(£■ )-5, 

 which are approximately distributed as normal (0,1) 

 random variables under the null hypothesis. Small val- 

 ues of X^^ indicate that the species is found less often 

 than predicted, whereas large values indicate it is found 

 more often. 



Results 



All species satisfied both the Cochran (1954) and Koehler 

 and Larntz (1980) criteria for the validity of the chi- 

 square test (Table 2). The null hypothesis that species 

 are randomly distributed among the crevice habitats was 

 rejected (P<0.0001) for all but one species, kelp greenling 

 (Hexagrammos decagrammus) (Table 2). 



We surveyed a total of 9804.1 m- of sea floor. Plat- 

 forms varied both in the amount of horizontal beam 

 exposed and, when exposed, the degree of gap between 

 beam and sea floor. At each platform, there was rela- 

 tively little annual variability in the amount of beam 

 exposed or the size and type of gap (Fig. 3). Mean size of 

 gap per platform over the entire study ranged from 2.5 

 (SD=1.1) to 0.5 m (SD = 0.4). (Fig. 3, Table 1). Platforms 

 Gail and Grace, in the east Santa Barbara Channel, had 

 the greatest amount of gap. In particular, almost none 

 of the bottommost beam at Gail was completely buried 

 and most of it was at least partially exposed. At the 



