Douglas et al : Geographic variation in cranial morphology of Stenella longirostns 



63 



greatest among-group variability relative to that 

 within groups and are used for more detailed com- 

 parisons with environmental variables (presented 

 below). 



As indicated in Table 4, the first canonical 

 variable is most influenced by Postorbital W. (Fig. 

 8A). In Figure 7, blocks that are large for this 

 character are to the right, while those that are 

 small are to the left. When considering only those 

 blocks with more than one specimen (i.e., those 

 shown with solid symbols in Fig. 7), the two blocks 

 from the vicinity of the the Hawaiian Islands 

 (0802 and 0702) are to the right, as are blocks 

 predominantly from the southern portion of the 

 range. The blocks with single specimens (which 

 tend to be more westerly) also are to the right. 

 Specimens from blocks to the north and east are 

 smaller; they are depicted to the left in Figure 

 7. Some west-central blocks group with the 

 southern blocks, while others are intermediate or 

 group with those to the northeast. The second 

 canonical variable contrasts blocks from the 

 Hawaiian Island area (0702, 0802) with the others 

 (see Fig. 7); in the characters reflected by this 

 variable, values of block 0812 (which is northern, 

 but to the east) show some similarities to those 

 for 0702 and 0802. 



Mantel test for geographic patterning 



Individual characters were evaluated with respect 

 to geographic patterning using Mantel tests, as 

 well as matrix correlations that compare inter- 

 block geographic distances (or reciprocals of these 

 distances) and character differences between localities. 

 Of the 30 characters, 66.7% (20) show statistically 

 significant regional patterning indicating that geo- 

 graphic distances (in nautical miles) and interblock 

 character differences are interrelated (Table 5). For 

 measures showing significant f -values the greatest 

 character differences tend to be between blocks that 

 are farthest away from each other, while nearer 

 localities are more similar. 



Local patterning, as indicated by a significant nega- 

 tive association of distance reciprocals and character 

 differences, was found in 83.3% (25) of the characters 

 (Table 5). All characters that showed regional pattern- 

 ing also exhibited local patterning. 



Principal component projections also were assessed 

 in terms of geographic patterning. As indicated at the 

 bottom of Table 5, component I (Fig. 4) has strong 

 regional and local patterning; component II has signifi- 

 cant local patterning. Canonical variables 1 (Fig. 8B) 

 and 2 both exhibit marked regional and local pattern- 

 ing (Table 5). 



Morphological-environmental covariation 



Figure 9 is a dendrogram indicating absolute correla- 

 tions among the 13 environmental variables, sub- 

 dividing them into five clusters. Sea Current (N., 

 Winter) is in a group by itself and quite different from 

 the others. Sea Current (W, Winter) and Oxygen Min. 

 Layer (Depth) are in the second cluster, which joins 

 with a group of five variables involving surface mea- 

 sures of temperature, oxygen, and salinity. The fourth 

 cluster involves two measures of solar insolation, and 

 the fifth reflects aspects of water depth. 



The loadings of environmental variables on the first 

 three environmental principal components are given 

 in Table 6. The first component statistically explains 

 33.0% of the total character variance, the second 

 23.2%, and the third 15.8% (cumulatively 72.0%). Maps 

 (Fig. 10) depict projections of the 25 blocks with two 

 or more 5. longirostris onto the first two environmen- 

 tal components. Environmental component I has rela- 

 tively high values for blocks between 5° and 15 °N, with 

 intermediate values to the north and low values south 

 of the Equator (Fig. lOA). Sea Surface Temp. (July) 



