634 



Fishery Bulletin 91|4). 1993 



10 



> 

 P 

 < 



JJI 



-10 



-■ i 



D [ 



DOLPHIN GROUPS 



 Northern Offshore Spotted 

 EH Whilebelly Spinner 

 I I Common 



"D 







"id "d[ 



10 



20 



10 



20 



10 



20 



25% Coverage 



50% Coverage 

 FLEET SIZE (5,10,20) 



75% Coverage 



Figure 5 



Average relative bias (%) in estimates of annual dolphin mortality as a 

 function of fleet size and percent observer coverage. Dashed line at +5"7r 

 and -5 f /r indicates management target. 



strongly but still noticeably by fleet size 

 (Figs. 10, 11, and 12). 



Mortality estimates were most stable and 

 confidence intervals dramatically reduced 

 for offshore spotted dolphins at high (75%) 

 observer coverage (Fig. 10). Mortality esti- 

 mates were more variable and confidence 

 intervals wide and ragged, even with a rela- 

 tively large fleet (20 boats), for offshore spot- 

 ted dolphins at low observer coverage (5%). 

 Intervals were most unstable at the combi- 

 nation of lowest observer coverage and 

 smallest fleet size 4 . Confidence interval pat- 

 terns for whitebelly spinner dolphins were 

 intermediate, being more variable than pat- 

 terns for northern offshore spotted dophins 

 but being less variable than patterns for 

 common dolphins (Fig. 11). At low coverage 

 and small fleet size, confidence intervals for 

 whitebelly spinner dophins showed the same 

 bimodality characteristic of confidence in- 

 tervals for common dolphins under all con- 

 ditions. At high coverage, confidence inter- 

 vals showed the same relatively stable and 



Distributions of mortality estimates for offshore spot- 

 ted dolphins tended to be relatively narrow and 

 unimodal for fleets of both 5 and 20 vessels, at high 

 coverage (75%; Fig. 7). Distributions remained 

 unimodel but were more dispersed, at low coverage 

 (25%). Distributions for whitebelly spinner 

 dolphins were also unimodal in general but 

 tended to be more dispersed than was the 

 case for offshore spotted dolphins (Fig. 8). 

 Distributions for common dolphins were 

 markedly bimodal and dispersed under all 

 sampling conditions, reflecting the selection 

 (or not) of the one trip with unusually high 

 kill (Fig. 9). 



Modal values of mortality estimates for 

 all three dolphin types increased with in- 

 creasing fleet size (not surprisingly, because 

 more boats generally kill more dolphins) re- 

 gardless of coverage level. Dispersion also 

 increased with fleet size, more obviously 

 when observer coverage was low than when 

 coverage was high, as more data became 

 available for analysis. 



Confidence limits 



Confidence limits were affected similarly to 

 frequency distributions. Limits were affected 

 most strongly by dolphin group type, very 

 noticeably by percent coverage, and less 



4 The irregular pattern of the confidence intervals observed for mortal- 

 ity estimates from small fleets (e.g., 5 boatsl with low coverage (e.g., 5 

 boats, 259f coverage! are not surprising because this combination of 

 fleet size and percent coverage means that the mortality estimates 

 are being derived from about 4 observed trips during an entire year 

 (assuming 5 boats make 3 trips/year, 15 trips X 0.25 = 4 trips). 



25% Coverage 



50% Coverage 

 FLEET SIZE (5,10,20) 



5 10 20 



75% Coverage 



Figure 6 



Coefficient of variation (percenti in estimates of annual dolphin mortality 

 as a function of fleet size and percent observer coverage. Dark horizontal 

 bar at 20 r r indicates management target. 



