170 



Fishery Bulletin 98(1) 



dard 2-m beam trawls in order to reduce the amount 

 of time the trawl took to reach the bottom and to 

 ensure that it was heavy enough to remain there. 

 To accommodate the increased weight (about 38 kg) 

 and prevent the trawl from digging into the sedi- 

 ment, the skids were also longer (0.85 m) and wider 

 (0.1 m). Typically, 2-m beam trawls are not used in 

 depths greater than seven meters (Rodgers, 1992; 

 Gibson, 1994; but see Pearcy, 1978); however, these 

 modifications enabled us to trawl easily in waters as 

 deep as 100 m. 



Fish samples were sorted on deck and all recently 

 settled juvenile fishes (age-0) were preserved in 957c 

 denatured ethanol. These fish were measured in the 

 laboratory to the nearest tenth of a millimeter standard 

 length (SL) with digital calipers. Older fish were mea- 

 sured on deck to the nearest millimeter and returned 

 to the water With the exception of Sebastes and Uro- 

 phycis spp., all age-0 fishes were identified to species. 

 Fish stages were identified according to morphologi- 

 cal changes such as squamation and, for flatfish, eye 

 migration, as well as analysis of length freo.uencies of 

 cohorts. The relative abundance and composition of 

 the remaining sample (shell hash, sand dollars, and 

 shrimp) were estimated from subsamples. 



Length-frequency distributions of each species col- 

 lected were calculated for each cruise. To account for 

 differences in sampling effort among cruises, abun- 

 dance data were standardized to the average cruise, 

 which had 63 tows (21 stations x 3 tows each) last- 

 ing five minutes each. Length frequencies were then 

 used to estimate the size range of the age-0 fishes 

 collected during each cruise, as well as to infer some 

 growth rates of the age-0 cohort between cruises. 



Distribution analysis 



Abundance data were standardized to the average 

 number offish per 1000 m'^. Because we were specifi- 

 cally interested in early juveniles, and adults were 

 not efficiently collected, we limited our analysis to 

 age-0 fishes. All species of age-0 fishes collected were 

 included in the analyses. The distribution of age-0 

 fishes was evaluated with respect to a suite of envi- 

 ronmental data (bottom temperature, bottom salin- 

 ity, and depth) by comparing weighted means for 

 each species (Scott, 1982). For a given species, an 

 environmental variable such as temperature was 

 weighted by the abundance of that species at each 

 sampling station. The sums of each weighted vari- 

 able were then divided by the total abundance of 

 that species collected to yield a mean value for that 

 species on that environmental parameter. 



Cross-shelf migrations with ontogeny are a common 

 feature of demersal fishes, particularly flatfish (Riley 



et al, 1981; Toole et al., 1997). To determine changes 

 in cross-shelf distribution with size, weighted mean 

 depths by size class were calculated for the more 

 abundant species collected. In particular, distinct dif- 

 ferences between th^ location of near-settlement size 

 fishes and larger age-0 fishes might indicate migra- 

 tion between settlement and nursery habitats. 



Environmental correlates 



Distribution and abundance of fishes in relation 

 to underlying multivariate environmental gradients 

 were analyzed by canonical correspondence analy- 

 sis (CCA) by using the software program CANOCO 

 (ter Braak, 1992). This analysis has been widely 

 applied in the field of community ecology (ter Braak, 

 1995; ter Braak and Verdonshot, 1995; Rakocinski et 

 al., 1996); it entails reciprocal averaging of species 

 and environmental data based on the assumption 

 of a unimodal response of species abundance to the 

 environment (Palmer, 1993; ter Braak, 1995). In a 

 comparison of ordination techniques. Palmer (1993) 

 found that CCA was robust, being less susceptible 

 to spurious results such as the "arch effect" often 

 common to principal components analysis (PCA). 

 Furthermore, Palmer's (1993) simulations of CCA 

 illustrated that noisy or skewed species data could 

 be compensated for, and a variety of data types and 

 sampling design were possible. 



In total, 25 environmental variables (Table 2) were 

 sampled during the cruises and included in the CCA 

 analysis. Bottom temperature and salinity (from CTD 

 casts), station depth (from an onboard depth sounder), 

 latitude and longitude (global positioning system 

 [GPS]), and distance from shore (nautical charts) were 

 employed as continuous measures during the analy- 

 sis. The relative abundances of nonfish constituents 

 collected by the trawl were also considered to be envi- 

 ronmental variables. Information on surficial sediment 

 character at each station, another environmental vari- 

 able we employed, was obtained from published data 

 from the Marine EcoSystems Atlas (MESA) program's 

 New York Bight Project (Freeland and Swift, 1978). 

 Although these data are independent of this study, the 

 MESA sampling area was the same and the spatial 

 resolution of its sediment sampling was fine enough 

 to obtain general information suitable for our analysis. 

 The data on nonfish trawl constituents and surficial sed- 

 iments were entered into the analysis as scaled values 

 (i.e. as a number describing relative abundance among 

 stations and tows). Because collections of both species 

 and environmental data were made year-round, sea- 

 sonal variables were also included and coded in the anal- 

 ysis as a set of nominal variables: spring (March-May), 

 summer (June- August), fall (September-November), 



