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



445 



Using the program Distance 3.5 (Research Unit for 

 Wildlife Population Assessment, University of St. An- 

 drews, UK), we fitted detection functions to perpendicu- 

 lar distance data to estimate ESW (note that this value 

 is derived directly from f(0)). Akaike's information crite- 

 rion (AIC) was used to select among models fitted to the 

 data. Models and adjustments were the following: haz- 

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

 normal/cosine, uniform/cosine (Buckland et al., 1993). 

 Following Buckland et al. (1993), perpendicular sighting 

 distances were truncated to eliminate the farthest 5% of 

 sightings and binned manually for /10) estimation. 



The coefficient of variation (CV) for the abundance es- 

 timate was calculated from the coefficients of variation 

 of each variable element in Equation 1 above (Buckland 

 et al., 1993): 



CV(N 5 )=JcV-{)]) + CV 2 (S) + CV 2 [ESW). 



(2) 



The CV(n) was estimated empirically as recommended 

 by Buckland et al. (1993): 



CV{n) = 



vardi) 



(3) 



(4) 



where var(«) = /.£/,<«, II,-n I L) : I ik-\). 



lj = the length of transect line i; 



nj = the number of sightings on transect i; and 



k - number of transect lines. 



CV(S) was estimated from the standard error of the 

 mean group size. CVlESW) was estimated with the 

 bootstrapping option in Distance 3.5 software. This 

 process incorporates uncertainty in model fitting and 

 model selection (Buckland et al., 1993). 



Measuring the effect of attraction 



Conventional line-transect estimates can be biased as 

 a result of responsive movement of the target species 

 and animals on or near the trackline being missed by 

 observers (Buckland et al., 1993). Buckland and Turnock 

 (1992) presented a method using co-ordinated boat and 

 helicopter surveys to quantify and adjust for the com- 

 bined effects of responsive movements of dolphins to the 

 boat and to eliminate the bias from observers failing to 

 see animals on or near the trackline. Their approach is 

 better suited to the restricted space available on small 

 boats than to a dual-platform approach (Palka and Ham- 

 mond, 2001). Additionally, sightings can be made much 

 farther ahead (reducing the possibility that the animals 

 have already responded), and the two sighting teams 

 are totally isolated from each other. For these reasons 

 we adapted Buckland and Turnock's (1992) approach in 

 our trials of 1998-99. 



Simultaneous boat-and-helicopter surveys were car- 

 ried out to the south of Banks Peninsula, predominantly 

 between Birdlings Flat and the mouth of the Rakaia 

 River. This area was chosen because it displayed rep- 

 resentative and varying densities. 



A Robinson R22 helicopter with pilot and one observer 

 (ES) followed a zig-zag flight path approximately 1.5 km 

 in front of the boat, traveling out to 1000 m on either 

 side of the vessel's trackline at a height of 500 ft ( 152 m) 

 (Fig. 4). To aid the process of tracking sightings from 

 the air, sighting positions were marked with Rhodamine 

 dye bombs. 2 The position of the helicopter in relation 



Dye bombs consisted of a tablespoon of Rhodamine dye in a 

 paper cup 2 /3 filled with sand. An additional (empty) paper 

 cup was taped upside down on top of the first cup with 

 paper-based masking tape. On impact the two cups broke 

 apart, releasing the sand+dye mix into the water. 



