628 



Fishery Bulletin 98(3) 



bottom at 0.4-0.9 knots. Transects were purposely of short 

 duration to maintain constant depth within the rock habi- 

 tat at each station. Each transect was documented con- 

 tinuously with a high 8-mm video camera and associated 

 lights that were externally mounted on the starboard side 

 of the submersible. The scientific observer verbally anno- 

 tated each videotape, identifying, counting, and estimat- 

 ing size of all fishes in front of the starboard viewing port. 

 A hand-held dive sonar was aimed at objects (e.g. large 

 fish and boulders) along the transects from inside the sub- 

 mersible to estimate distance from the observer to the 

 object; this procedure helped us to estimate the width of 

 the transect. After each dive, divers transcribed observa- 

 tions on fishes and habitat from video tapes into a comput- 

 erized database on board the support vessel. 



Two parallel lasers were mounted on either side of the 

 external video camera at the fixed distance of 39.5 or 20.0 

 cm apart, in association with different laser systems each 

 year. The laser spots were projected onto the seafloor habi- 

 tats. They were visible to the observer, recorded onto the 

 video tape, and were critical in accurately estimating the 

 size of fishes, distance traveled along a transect and area 

 of habitat patches. We made measurements by comparing 

 the size of a fish or habitat feature to the known spacing of 

 the two bright laser spots when the object was perpendicu- 

 lar to the camera and lasers (Tusting and Davis, 1993). We 

 estimated the length of each transect, independent of sub- 

 mersible speed and bottom currents and type, by counting 

 the number of laser-spot intervals as they moved along the 

 substrata in the video transect (much like using a yard- 

 stick, end-over-end along the transect). 



Microhabitat of each fish within the transect was char- 

 acterized from the video tapes. Various combinations of 

 substratum type, including mud, pebble, cobble, boulders, 

 and rock ridge (see Greene et al., 1999, for definitions), 

 were categorized according to primary (at least 50'/( of 

 the area viewed) and secondary (>20'^ of the area viewed) 

 microhabitat (conforming to Stein et al.. 1992). Relief was 

 categorized as flat (0-5°), low (5-30° I, and high (>30°). 

 Each surface area of uniform habitat (i.e. a patch) along 

 the quantitative transect was measured to the nearest 

 0.1 m^. Species-specific abundance was standardized per 

 area of associated habitat patch. The habitat patch was 

 used as our sample unit. 



Data analyses 



Similarity of assemblages of nonschooling fishes among 

 the different combinations of substrata was evaluated 

 with cluster analysis on the basis of abundance of each 

 species standardized by area of associated bottom type 

 in each patch. Only species representing >V'/i of the total 

 abundance in each bottom type category and only bottom 

 types representing >l'7i of the total area surveyed were 

 used in this analysis. Only nonschooling (i.e. nonpolar- 

 ized aggregations or solitary individuals) benthic fishes 

 were included in our analyses because schooling fishes 

 commonly were more abundant in midwater above our 

 field of view and therefore could not be accurately enumer- 

 ated. Clustering was performed with the average linkage 



method and with Euclidean distance as a measure of dis- 

 similarity (SYSTAT, 1992). Dissimilarity among clusters 

 >50'7f of the maximum overall distance was considered a 

 major division and used to define distinct habitat guilds of 

 fishes iserisu Root, 1967). 



Further analyses were focused on nonschooling fish spe- 

 cies that dominated the rock habitat guild, as defined by 

 the cluster analysis. These are some of the species impor- 

 tant in commercial and recreational catches (Weinberg, 

 1994; Mason, 1995, 1998). We used the incidence of fish- 

 ing gear and associated debris, observed on the seafloor 

 during the quantitative fish transects, as a relative index 

 of fishing activity throughout our study area. Statistical 

 differences in abundance (number of fish per 100 m^ of 

 habitat patch) of those species in the rock habitat guild 

 were analyzed among five sites of varying fishing activity 

 by using analysis of variance (AN OVA) with equal sample 

 variances and otherwise by resampling statistics (Bruce et 

 al., 1995). We used Cochran's test for homogeneity of vari- 

 ance (Winer, 1971). 



Differences in size of selected dominant species were 

 tested among two arbitrarily chosen 100-m depth catego- 

 ries (i.e. shallow [75-175 m] and deep [176-275 ml) and 

 the five sites were tested by using two-factor AN OVA (with 

 homogeneity of sample variances) where appropriate. 



Overall species diversity was calculated as 



H' 



^[pjllnpj 



where s = number of species; and 



p, = proportional abundance of species /. 



Richness (number of species), and evenness (J' = H' IH\^f^), 

 as well as species diversity, were evaluated for all habitat 

 types ( see Krebs, 1989) and then among sites just within the 

 rock habitat guild in shallow and deep water Sufficiency 

 in the number of samples necessary to reliably charac- 

 terize overall diversity for each habitat type was exam- 

 ined by plotting cumulative numbers of species against 

 the sample unit (both for number of patches and area of 

 habitat surveyed). These plots indicated that the number 

 of samples was sufficient to yield a reliable estimate of 

 diversity for comparisons among all habitats and for com- 

 parisons among sites in shallow and deep rock habitat (i.e. 

 the number of samples evaluated for diversity always sur- 

 passed the number comprising 95% of the species; see data 

 on Figs. 8 and 9). 



Results 



Geophysical mapping of habitats 



A physiographic representation of the relatively high-res- 

 olution bathymetric data (Fig. IB; production assisted by 

 the U.S. Geological Sui-vey, Menlo Park, CA) helped us to 

 visualize canyon morphology, to identify areas of high relief 

 and potential slumping, and to select submersible dive 



