Dawson et al.: Line-transect surveys of Cephalorhynchus hectori 



447 



A correction factor for abundance estimates of Hector's 

 dolphin groups can be estimated by 



c = D L ID,. 



(9) 



Using Distance 3.5, we fitted a half-normal model with 

 cosine adjustments to estimate /10). The half-normal 

 model was fitted to helicopter data to estimate /",,(0) and 

 the uniform model with cosine adjustments was used to 

 estimate f hs (0)- All were selected by using AIC. Potential 

 model choices were the following: hazard/cosine, hazard/ 

 polynomial, half-normal/cosine, half-normal/hermite 

 and uniform/cosine (Buckland et al., 1993). Truncation 

 distance was 640 m for boat sightings, and 1000 m for 

 helicopter and duplicate sightings. To ensure that only 

 high-quality data were used to estimate effective half 

 search widths, sightings for which range (radial distance) 

 was estimated by eye and those made during Beaufort 

 sea state >2 were removed before f(0) estimation. 



The error for the correction factor (c) was estimated 

 by bootstrapping on transect lines and applying the 

 estimation procedure to each of 199 bootstrap data sets. 

 The standard deviation of the bootstrap estimates was 

 used as the standard error of c. 



Ideally, the correction factor would be estimated sepa- 

 rately for each survey from separate sets of boat-and-he- 

 licopter trials conducted in areas of representative den- 

 sity. Financial and logistical constraints prevented this; 

 therefore the correction factor estimated in 1998-99 was 

 applied to each of the line-transect surveys reported in 

 the present study. We note that this is not uncommon 

 (e.g., Carretta et al., 2001). 



Unbiased abundance estimates were calculated by 



N, 



"■N, 



(10) 



The CVs of the corrected abundance estimates (N L ,) 

 were calculated with the following equation (Turnock 

 et al„ 1995): 



CV(N U ) = JCV 2 (£) + CV 2 (N S ), 



where CV(c) 



5£(r) 



(11) 



(12) 



Upper (N uc ) and lower (N LC ) 95% confidence inter- 

 vals for N v were calculated by using the Satterthwaite 

 degrees of freedom procedure outlined in Buckland et al. 

 (1993). This procedure assumes a log-normal distribu- 

 tion of N c , using 



N LC = N L . I C, and 

 N uc = N U C, 



where C = exp \ r, ( 0.025 ) log, 1 + 



-[CV(N, >]" 



(13) 



(14) 



The Satterthwaite degrees of freedom (df) for corrected 

 abundance estimate confidence intervals were calcu- 

 lated by 



df= 



CV\N, ) 



Cl'V) CV\N S ) 



(15) 



tf-1 



df s 



where B is the number of bootstrap samples, and df s is 

 the Satterthwaite degrees of freedom for the uncorrected 

 abundance estimate, N s (see Buckland et al., 1993). 



The CV of combined abundance estimates (N ai ) was 

 computed by 



SEUouih 



J{SE\N m 



) + SE-(N,,) + ... + SE-iN. 



«)}• 



and 



SEiloiah 

 CV{total) = -^- 



N, (total) 



(16) 



(17) 



Results 



The three line-transect surveys covered 2061 km of tran- 

 sect, and 231 sightings were used to estimate density 

 (Table 2). Sighting rates were highest around Banks 

 Peninsula (Table 3). 



The simultaneous boat-and-helicopter surveys indi- 

 cated that boat observers missed 11.4% of the dolphins 

 on the trackline, but that strong responsive movement 

 towards the boat resulted in apparent densities twice as 

 high as they normally would be (Table 3). If the observ- 

 ers' attention was drawn to dolphin groups by the posi- 

 tion of the helicopter, the results of these trials would 

 be biased. This is unlikely, however, because several 

 groups sighted by the helicopter observer subsequently 

 passed within 200 m of the boat and were not seen by 

 observers. We saw no evidence that the dolphins were 

 affected by the helicopter. 



Detection functions for boat-and-helicopter sightings 

 (Fig. 5, C and D) are relatively smooth in comparison 



