Fiedler and Reilly: Interannual variability of dolphin habitats and abundances 



453 



ciprocal of the distance from the grid point. The 

 range of observations around a grid point was in- 

 creased in increments of 0.4 degrees latitude and 2 

 degrees longitude to obtain a minimum sample size 

 of 20 for each grid point. Thus, local grid resolution 

 decreases in data-poor regions, generally south of 

 the equator where the maximum distance required 

 was up to 20 degrees longitude. Within the MOPS 

 area, sufficient observations were available within 

 2 degrees latitude and 10 degrees longitude of 71% 

 of the gridpoints, and within 4 degrees latitude and 

 20 degrees longitude of 95% of the gridpoints. We 

 converted observations to anomalies (deviations 

 from the seasonal mean) before gridding to reduce 

 the spatial variability of the observations. This mini- 

 mized bias caused by interpolation over or extrapo- 

 lation into large data gaps. 



Relationships among abundances of dolphin spe- 

 cies and environmental variables were analyzed by 

 using CCA as described in Reilly and Fiedler (1994). 

 Encounter rate, equal to number of schools sighted 

 per unit of sighting effort (trackline distance), was 

 used as a measure of relative abundance. The final 

 abundance estimate also depends on school size and 

 effective track width. However, Reilly and Fiedler 

 (1994) found that weighting encounter rates by es- 

 timated school size in the CCA produced essentially 

 the same species-environment patterns. Therefore, 

 schools of all sizes occupied approximately the same 

 habitats and school size variability within these habi- 

 tats was not related to the environmental variables 

 included in the analysis. 



CCA was performed on MOPS sightings and en- 

 vironmental data as in Reilly and Fiedler ( 1994), ex- 

 cept that mixed schools of spotted and spinner dol- 

 phins were counted as schools of both species rather 

 than as an additional "species." Also, we used Hill's 

 symmetric scaling of species and site scores (S=-3 

 in our implementation of CANOCO 2 ). This alterna- 

 tive scaling of the ordination had no qualitative ef- 

 fect on species-environment patterns but seemed to 

 give more reasonable results at the edges of the 

 study area when scores were combined to quantify 

 species habitat distributions. 



Habitat quality for species i (H t ) at a gridpoint 

 was calculated from the Gaussian responses fit to 

 the two dominant canonical axes by CCA. The re- 

 sponse to each environmental axis was calculated as 

 a normal probability density function: 



2 Ter Braak, C. J. F. (1988). CANOCO— a FORTRAN program 

 for canonical community ordination by I partial I [detrendedl 

 [canonical] correspondence analysis, principal components 

 analysis and redundancy analysis (version 2.1). Tech. Rept. 

 LWA-88-02, Groep Landbouwwiskunde, Postbus 100, 6700 AC 

 Wageningen. The Netherlands. 



H tJ = t-j 1 exp (^ -0.5 * ((xj - «y )/ty) J , 



where, 



x = the site (gridpoint) score on environmental 



axis./', 

 u t j = the species i score (optimum) on axis j, 

 t u = the tolerance (standard deviation) of species 



i on axis j. 



Species scores and tolerances (u- and £••) were 

 output by CANOCO as part of the CCA. Site scores 

 (x ) were calculated as linear combinations, defined 

 by the output canonical axis coefficients, of normal- 

 ized and gridded environmental observations. Habi- 

 tat quality, H t , was then calculated as the geomet- 

 ric mean of HJ&; each Hy was scaled so that the 

 mean value is 1.0 during 1975-90. Thus, H i is equal 

 to the abundance expected at a site, based on local 

 environmental conditions, divided by the mean 

 abundance in the study area during 1975-90. 



Point estimates of annual abundance were pro- 

 vided by Anganuzzi 3 for pooled stocks: spotted dol- 

 phins include northern and southern offshore spot- 

 ted dolphins, whitebelly spinner dolphins include 

 northern and southern whitebelly spinner dolphins, 

 and common dolphins include northern, central, and 

 southern common dolphins. No estimates were made 

 for striped dolphins, which were rarely set on by 

 tuna vessels. 



Results 



The species-environment biplot (Fig. 1) summarizes 

 the results of the CCA of five species and three en- 

 vironmental variables observed during 1986-90 

 MOPS research surveys. The eigenvalues of the 

 three canonical axes were 0.296, 0.074, and 0.001. 

 The first two axes explained 99.7% of the species- 

 environment variance accounted for by all three 

 axes. Therefore, the third axis was not used. The 

 first two axes explained 20.5% of the total variance 

 of species encounter rates (Table 2). 



Positive scores on canonical axis 1 indicate cool 

 surface temperature and a shallow, weak ther- 

 mocline (Table 3). These are characteristics of the 

 productive "cool upwelling" habitat that we identi- 

 fied with the first axis in the complete CCA (seven 

 species and six environmental variables, Reilly and 

 Fiedler, 1994). This habitat is found in the equato- 

 rial and eastern boundary current (Peru and Cali- 



3 Anganuzzi, A. Inter-American Tropical Tuna Commission, 8604 

 La Jolla Shores Drive, La Jolla, CA 92038. Pers. commun. De- 

 cember 1991. 



