442 



Fishery Bulletin 92(2). 1994 



30 



S 20 



10 - 



40 



30 



10 



ii 



temperature. These are character- 

 istics of "coastal tropical" habitat 

 found along the coast of Central 

 America, where the surface layer is 

 more stratified and upwelling is 

 more intermittent and localized 

 than in the cool upwelling habitat. 

 Whitebelly spinners, alone and with 

 spotted dolphins, had large negative 

 axis-2 scores, while eastern spin- 

 ners with spotted dolphins had posi- 

 tive scores. There was a strong sepa- 

 ration on axis 2 between mixed 

 schools of whitebelly spinners with 

 spotted dolphins, and schools of 

 eastern spinners with spotted dol- 

 phins. Schools of spotted dolphins 

 alone had near-zero axis 2 scores. 

 Striped dolphins loaded near the 

 origin of both axes. 



The spatial distributions of yearly 

 axis 1 scores are mapped in Figure 



6. Areas with positive scores are 

 shaded to allow quick appraisal of 

 changes between years (interpreted 

 below). Also plotted in Figure 6 are 

 sighting localities for spotted and 

 common dolphins. Spotted dolphins 

 occurred mostly in negative areas, 

 common dolphins in positive areas, 

 but with some overlap. 



Spatial distributions of yearly 

 axis 2 scores are mapped in Figure 



7, with positive scores shaded. 

 Whitebelly spinners occurred al- 

 most exclusively in waters with 

 negative axis-2 scores (Figs. 5 and 

 7). Eastern spinners ranged 

 throughout both positive and nega- 

 tive areas; a modest majority were 

 found in positive areas. They were 

 less closely associated with this axis 

 than whitebelly spinners (Fig. 5) and seemed to be 

 found in the eastern (more coastal) part of the warm 

 tropical habitat defined by negative axis-1 scores. 



Interannual variability 



We obtained only a slight increase in the percent of 

 variance explained for the dolphin data (14.7% to 

 15.1%, Table 5) from addition of categorical variables 

 representing the five sampled years, in addition to 

 the six oceanographic variables. An ordination biplot 

 from this analysis (Fig. 8) shows that the centroid 

 for 1988 (year 3) loads farthest from the origin. Its 



T 1 1 1 1 1 1 1 1 I ' I I - 



llL. 



-I 1 1 1 1 1 1 1 1 1 1 1 1 - 



A 



OjjJ* j&j&> Kp<&> 0&<y1* -»°\3fr * 6 r** iP-AP ^°&* X&s&> <&&*> O&tfl* ^ ,3& \t tfb oD ,/jP 



40 



l — i — l — l — l — l — l — i — i — i — i—i — i — r- 



lllW 



- 10 - 



-I 1 1 1 1 1 1 1 1 1 1 T - 



_J 1 J 1_ 



lO^Tft j&j&t oP<pt> O&ftT* ^il* ^\1* tP'AP ^■* D ^ 1 * -0»c3 6 OPqZ 6 O^tf* ^ \l b ^ \1 6 1? JP 



\ii> 



% 20 

 O 



T — I — I — I — i — i — I — l — i — r- 



.ulllllll 



Qj&* -0&o3* &<& 0»oJ« *0 ,20 ^,1* 10^0 „^°.Ol* J&X& 0°02 6 O&tf 6 \0 ,j6 \fi \jt tfi j& 



Figure 4 (Continued) 



location represents the cooler, more productive con- 

 ditions associated with the 1988 La Nina. 



An analysis including just years as categories, 

 without oceanographic variables, explained only 

 about 2% of the dolphin variance, but the dominant 

 eigenvalue and trace were both significantly differ- 

 ent from zero (Monte Carlo P-values=0.01 and 0.02, 

 respectively). After extraction of the variance asso- 

 ciated with the six oceanographic variables (by de- 

 fining them as covariables) the ordination was not 

 significant (Monte Carlo P=0.058, Table 5). 



