176 



Fishery Bulletin 98(1) 



rachnius parma ) and valves of surf clams (Spisula 

 solidissima) dominated the midshelf group (40-70 

 m) of benthic organisms. Shell hash collected mid- 

 shelf consisted of fine fragments of sand dollar tests 

 and larger pieces of clam valves. Small, recently 

 settled rock crabs (Cancer borealis) were found in 

 high numbers (>100 per 5-min trawl) during the 

 summer in this type of sample. Margined sea stars 

 (Pontaster tenuispinus) and sea scallops iPlacopec- 

 ten magellanicus) made up most of the trawl catch 

 for deeper outer-shelf stations (70-90 m). Young 

 pandalid shrimp (Pandalus montagui) constituted a 

 large fraction (30-90%) of samples from these sta- 

 tions in late summer and fall. Larger crustaceans 

 such as American lobsters (Homarus americanus) 

 and large rock crabs (C. borealis) were occasionally 

 abundant (>10 per 5-min trawl) at these stations. 

 Several were collected during each cruise. 



Other groups of macrobenthos were ubiquitous. 

 Hermit crabs and cancer crabs of intermediate sizes 

 were found throughout the stations sampled. Although 

 they dominated the midshelf, sand dollars were also 

 collected at inner-shelf stations. Other large benthic 

 fauna such as horseshoe crabs (Limulus polyphemus) 

 and spider crabs (Libinia emarginata) were collected, 

 but these collections were sporadic. 



Species assemblages and environmental variables 



Four canonical axes, each representing a linear com- 

 bination of the environmental data, were calculated 

 from the data set. These four axes together accounted 

 for 36.7% of the variation in species abundance and 

 82.1% of the cumulative variation, in relation to the 

 total variation explained by the environmental vari- 

 ables. A summary of eigenvalues and the variance 

 accounted for by each axis is given in Table 6A. The 

 variance explained by the entire ordination, as well 

 as the first axis, was more significant than expected 

 by chance, as calculated by a Monte Carlo permuta- 

 tion test («=99 iterations, P=0.01 for both tests). 



Interpretation of the relationships between envi- 

 ronmental variables and the CCA axes involves deter- 

 mining which variables are most correlated to the 

 axes. One intuitive and effective method of accom- 

 plishing this is to examine an ordination plot of envi- 

 ronmental variables (Fig. 9). Environmental variables 

 with large components along a CCA axis have high 

 correlations to that axis. However, the results of the 

 ordination with 25 environmental variables were dif- 

 ficult to interpret because of inherent covariability 

 of variables with one another (Fig. 9, A-B). 



Forward selection of environmental variables 

 resulted in selection of five environmental variables 

 (bottom temperature, depth, the relative abundance 

 of scallops, longitude, and the relative abundance 

 of margined sea stars). Eigenvalues of the forward 

 selection of these five variables (Table 6B) are pre- 

 dictably lower than for the CCA with all 25 vari- 

 ables included (Table 6A). However, 59% of the 

 variability explained with all 25 variables included 

 was explained by these five environmental variables 

 alone. Temperature represented 23% of the entire 

 variability explained by the environmental data, 

 depth contributed 18%, scallop abundance explained 

 8%, and longitude and margined sea star abundance 

 combined explain an additional 5%. The remaining 

 variables, which did not add significantly to the 

 resulting explanation, each explained less than 4% 

 of the variation. Because only five variables were 

 selected, the four synthetic gradients (CCA axes) are 

 each highly correlated with one of the environmental 

 variables (Table 6A). Thus, to some extent each axis 



