452 



Fishery Bulletin 92(2). 1994 



for the Monitoring of Porpoise Stocks (MOPS) pro- 

 gram during August-November in 1986 through 

 1990 (Wade and Gerrodette, 1992). The MOPS sur- 

 veys were designed to cover the SOPS spotted, spin- 

 ner, and common dolphin stock boundaries. Reilly 

 and Fiedler (1994) showed that sighting rates of dol- 

 phin species or school types on the MOPS surveys 

 were related to concurrently measured environmen- 

 tal variables. We used a robust and efficient multi- 

 variate technique, canonical correspondence analy- 

 sis, to examine relations between spatial distribu- 

 tions of dolphin species and environmental vari- 

 ables. 



Canonical correspondence analysis (CCA) was 

 developed to relate community composition to 

 known variation in the environment. It is a form of 

 direct gradient analysis that directly estimates or- 

 dination axes as linear combinations of observed 

 environmental variables (Ter Braak, 1986). The 

 advantages of CCA for multivariate species-environ- 

 ment analyses and details of the method are dis- 

 cussed by Reilly and Fiedler (1994). CCA estimates 

 unimodal (Gaussian) responses of species along the 

 ordination axes. In general, the response is observed 

 abundance or probability of occurrence. We assume 

 that a species response (abundance observed at a 

 site in time and space) reflects the suitability of 

 environmental conditions at that site relative to the 

 species' optimal habitat or niche. This suitability, or 

 habitat quality, is defined by the response distribu- 

 tion along the axes. An observed response will also 

 include error caused by behavioral responses to the 

 environment that affect the detectability of schools. 



In this paper, we analyze relations between abun- 

 dance estimates and habitat quality. It must be 

 stated at the outset that in a 15-year record of popu- 

 lations with growth rates of only -0.02 yr _1 (Reilly 

 and Barlow, 1986), we are able to detect short-term 

 environmental effects on sampling but not long-term 

 effects on population size. Nevertheless, we demon- 

 strate that some of the interannual variability in 

 estimated abundances of ETP dolphins can be ex- 

 plained by environmental factors. 



Methods 



We used archived bathythermograph data to quan- 

 tify variability of surface temperature, thermocline 

 depth, and thermocline thickness in the MOPS 

 study area since 1975. These variables were shown 

 to be important in explaining variations in encoun- 

 ter rates in the MOPS surveys (Reilly and Fiedler, 

 1994). Other important variables (salinity and chlo- 

 rophyll concentration) have not been routinely ob- 



served with sufficient frequency to be used in this 

 historical analysis. Seasonal fields (gridded values) 

 of surface temperature, thermocline depth, and ther- 

 mocline thickness for the period 1975-90 were de- 

 rived from a bathythermograph data base originally 

 described by Fiedler (1992) and augmented with 

 data from other sources for this study (Table 1). 

 Thermocline depth is defined as the depth of the 

 20°C isotherm. Thermocline thickness is defined as 

 the difference in depth between the 20°C and 15°C 

 isotherms. 



Data were objectively gridded by seasons (Decem- 

 ber-February, March-May, June-August, Septem- 

 ber-November from 1975 through 1990) on a 2-de- 

 gree latitude-longitude grid from lat. 20°S to 30°N 

 latitude and from the coast out to long. 160°W. 

 Decorrelation scales, the distances required for a 

 substantial change in surface temperature or ther- 

 mocline depth, have been estimated as 3 degrees 

 latitude and 15 degrees longitude in this region 

 (Sprintall and Meyers, 1991). At each grid point, 

 means of at least 20 observations within up to 4 

 degrees latitude and 20 degrees longitude were cal- 

 culated. The observations were weighted by the re- 



Table 1 



Numbers of bathythermograph profiles, after 

 screening for errors and replicates, used to define 

 habitat quality in yearly seasonal grids ( 1975-90) 

 and in climatologies (1960-91). NODC = NOAA/ 

 NESDIS/National Oceanographic Data Center 

 CD-ROM NODC-03: Global Ocean Temperature 

 and Salinity Profiles, vol. 2, Pacific Ocean; 

 MOODS = Navy Master Oceanographic Observa- 

 tions Data Set, including non-NODC observations 

 through 1983 obtained from the Naval Oceano- 

 graphic Office through NODC and 1985-90 obser- 

 vations obtained from Steve Pazan, Scripps Insti- 

 tution of Oceanography; SOP = French-American 

 ship-of-opportunity observations obtained from 

 NOAA/ERL/Pacific Marine Environmental Labo- 

 ratory (Kessler, 1990); FSFRL = Japanese Far 

 Seas Fisheries Research Laboratory MBT data 

 obtained from PMEL and from NOS/Ocean Appli- 

 cations Branch (these data will be added to the 

 NODC data set in the near future). 



1975-90 



1960-91 



NODC 

 MOODS 

 SOP 

 FSFRL 



Total 



61,486 



10,741 



2,859 



2,350 



77,436 



127,365 



15,077 



11,305 



4,744 



158,491 



