Reilly and Fiedler. Interannual variability of dolphin habitats 



447 



Other, perhaps more sophisticated approaches are 

 possible. We present these only as examples. The 

 most straightforward approach, involving minimal 

 assumptions, would be to post-stratify the data for 

 each year separately, based on the spatial distribu- 

 tion of CCA axis scores and the weighted mean and 

 standard deviation of those scores for the species of 

 interest. This would be done to improve precision of 

 abundance estimates. Populations that have similar 

 means and standard deviations could use common 

 strata. For example, separate strata could be defined 

 by using axis 1 for common and spotted dolphins. 



CANONICAL AXIS 2 



130 120 1 10 100 »0 80 



• White-belly Spinner dolphin ^Eastern Spinner dolphin 



Figure 7 



Maps of distribution of canonical axis 2 for 1986-90. Nega- 

 tive areas are shaded. Eastern spinner dolphin, Stenella 

 longirostris, sighting localities are represented by closed 

 circles, whitebelly spinner dolphin, S. longirostris, localities 

 by open triangles. 



Axis 2 could be used to provide strata for whitebelly 

 spinner dolphins. Because we have probability dis- 

 tributions for the occurrence of these species along 

 the canonical axes, we would not be limited to use 

 just two strata but could use three or four. After the 

 data were stratified based on the species annual habi- 

 tat distributions, standard line transect methodol- 

 ogy would be followed. This is generically similar to 

 the post-stratification approach taken by Anganuzzi 

 and Buckland (1989) to reduce bias in estimates of 

 dolphin abundance from tuna vessel observer data. 

 A second possible approach, aimed at improving 

 accuracy, would quantify the amount of habi- 

 tat available within the study area each year, 

 for each population. The simplest quantifica- 

 tion scheme would define only two strata for 

 each. The cut-point between strata could be the 

 95% limit of the population's distribution on the 

 axis, or, less conservatively, the appropriate 

 upper or lower quartile. More complex schemes 

 using more than two strata could be developed, 

 as with the post-stratification, based on addi- 

 tional information in the species probability dis- 

 tributions. The amount of any stratum avail- 

 able in a year could be quantified by, say, lightly 

 smoothing and interpolating the CCA site 

 scores (to provide values for all locations) and 

 "sampling" the distribution with the actual 

 cruise tracks for the year. If for example com- 

 mon dolphin habitat was to be defined as axis 

 1 > [some value], the amount of ocean sampled 

 with axis 1 > [some value] in 1986 could be 

 scaled as 1.0. The amount sampled in subse- 

 quent years could be scaled to the 1986 amount. 

 The result would be a vector of values repre- 

 senting the amount of common dolphin habitat 

 available within the ETP by year. This vector 

 could then be applied to the encounter rate por- 

 tion of the line transect abundance estimate for 

 each year to account for changing availability 

 of common dolphin habitat. If interannual dif- 

 ferences were subsequently observed in the line 

 transect abundance estimates, we could be more 

 confident that they represent real changes in 

 abundance, rather than just apparent changes 

 due to spatial redistribution relative to sam- 

 pling effort following habitat shifts. 



In a separate study (Fiedler and Reilly, 1994) 

 we applied the CCA ordination approach devel- 

 oped here to investigate interannual variabil- 

 ity in abundance indices for ETP dolphins esti- 

 mated from tuna vessel observer data. We cal- 

 culated annual indices of habitat quality for 

 each dolphin species targeted by the tuna fish- 

 ery, for the years 1975-90, then compared these 



