Felley and Vecchione. Nekton habitat on the continental slope of North Carolina 



269 



FACTOR 



SpcA Lzdf Cncb 



Octo 

 Plsp Find 



Sqd Elpt Shmp 

 Hake Eel Rati Scr P 



Large anemones 



Small anemones. Tubes 



EPIFAUNA 



SpcA 



Octo 

 Sqd Scrp 



Cncb Shmp 



Find Plsp 



Elpt Rati Hake Lzdf 



Eel 



Few 

 sargassum debris present 



Cncb 

 Sqd Shmp 



r~ 



MOUNDS. CRINOIDS 



Many 

 sargassum debris rare 



Eel 



SpcA Octo 

 Plsp Elpt Find 



Rati Hake Scrp Lzdf 



Sparse 



HOLES, GRASS DETRITUS 



Dense 



-2 2 



Figure 1 



Distribution of species scores on axes representing habitat use by nekton species at 

 Cape Hatteras. Factor 1 represented habitat use according to different epifaunal as- 

 semblages, factor 2 according to density of mounds and crinoids, and factor 3 according 

 to density of holes and presence or absence of grass detritus. Species abbreviations are 

 as follows: Cncb=cancroid crab, Elpt=eelpout, Flnd=flounder, Lzdf=lizardfish, 

 Octo=octopod, Plsp=sergestid shrimp, Ratl=rattail, Scrp=scorpaenid, Shmp=shrimp, 

 SpcA=Species A (possibly offshore hake), Sqd=shortfin squid. 



poorly known deep-ocean nekton assemblage. Accept- 

 ing the assumption discussed above, we confronted 

 the following methodological questions: 1) which spe- 

 cies to analyze, 2) which environmental variables to 

 measure, and 3) at what scale to sample. Our an- 

 swers to these questions were pragmatic. We quan- 

 tified those forms we felt could be recognized easily 

 and consistently and were consistently visible on the 

 videotape. We measured as many variables as we 

 could quantify visually in an accurate and repeat- 

 able fashion and also included those variables that 

 seemed to be dominant in the environment (e.g. small 

 anemones, holes). We sampled at a spatial scale as- 

 sumed to approximate the area monitored by spe- 

 cies of the assemblage. Although the area viewed in 

 one minute of submersible cruising (a distance of ca. 

 15-30 m) is doubtless greater than the area moni- 

 tored by an individual at any particular moment, we 

 felt this to be the smallest manageable sampling unit. 

 Shorter intervals (e.g. 30 seconds) required an inor- 

 dinate amount of videotape stopping and starting. 

 Preliminary analyses suggested that increasing the 

 size of sampling units would create problems, includ- 



ing 1) a much smaller number of samples for the 

 analysis, 2) loss of information in 1/0 coded variables, 

 as these tended to become 1 (environmental attribute 

 present) in all intervals, and 3) loss of information in 

 variables whose scale of change was smaller than the 

 sampling unit. In each case, increasing the size of sam- 

 pling units tended to decrease the measurable associa- 

 tion between a species and particular environmental 

 variables (see also Schneider et al., 1987). 



Correlations between species occurrence and par- 

 ticular environmental variables were subjected to 

 multivariate analysis to find the patterns of habitat 

 use in this species assemblage. In accordance with 

 James and McCulloch's ( 1990) caveats, we recognize 

 and stress the correlational aspect of this study. Our 

 analysis produced artificial axes that only reflect real 

 trends. These artificial axes provided the data for 

 all further statistical tests. However, our interpre- 

 tations are strictly tied to the real variables repre- 

 sented by the artificial axes. When a factor had high 

 loadings for holes and small anemones, we inferred 

 that one of these variables, or some other variable 

 strongly related to them, was in fact affecting distri- 



