Biological Oceanography

(^) The result looks like the species numbers map (Fig. 14.4). That is, the contour of 60
species that appears in the east is attributable to a set of forms that cause identification
of a station cluster. Note that the cluster patterns are very much parallel to the depth
contours, but that the A cluster extends across the contours in the west. This implies
that, while depth is an important habitat determinant, something else is also important.
(^) Bilyard also applied a canonical correlation analysis, an ordination technique (Box
14.1) involving orthogonal axes. Clusters identified (Fig. 14.7) were the same as those
from CLUSB. That is usual. If the clustering is strong, i.e. if data really indicate
clusters, then most methods will show that. If it is weak, then different methods will
show different clustering. There are now methods for developing a probabilistic
evaluation of the “reality” of the clusters (Box 14.1). Bilyard’s clusters are tight
enough, distinctive enough that such testing would be superfluous. The only data
available to Bilyard for seeking habitat variables that might explain the cross-isobath
distribution of his A-cluster stations were some sediment-particle size analyses. He
plotted the positions of the stations, identified by their cluster designations, on a
triangle or “ternary” diagram of the proportions of clay, silt, and sand–gravel (Fig.
14.8). Cluster B, with two exceptional stations, is a group of stations with a high
proportion of coarse sediment. The two exceptional stations contained some gravel,
possibly ice rafted, which may be the habitat requirement of the group. Bilyard’s
conclusion is not unusual: sediment character is frequently a controlling habitat
variable for the composition of benthic fauna. On the other hand, Cluster B appears
from the map (Fig. 14.6) to be reasonably well explained by depth alone, since there
is an onshore–offshore gradient in sediment, coarser sediment settling closer to the
coast than finer. However, the isobath-crossing pattern of Cluster A is not particularly
well explained by sediment characteristics. Cluster A remains unexplained, a potential