340 



Fishery Bulletin 105(3) 



^.1 i=i '^ij hj 



(3) 



where k = the total number of transects within geo- 

 graphic stratum j; 

 t^j = the length (in km) of the ith transect in stra- 

 tum J; 

 T^- = the total length of all transects in stratum^; 

 n = the number of turtles seen on transect / in 

 stratum 7; and 



L- = the actual distance flown on transect i within 



y 



stratum J. 



The probability of detecting a leatherback turtle at 

 zero perpendicular distance, giO), is primarily influ- 

 enced by the proportion of time a turtle is unavailable 

 to be seen by the aerial survey team because it is diving 

 (availability bias; Marsh and Sinclair, 1989). In other 

 cases, animals may be present at or near the surface, 

 but missed by observers for other reasons, such as fa- 

 tigue or poor viewing conditions (perception bias). In 

 this study, no correction was available for perception 

 bias. Availability bias was estimated from leatherback 

 turtle dive data and the estimated visible depth range 

 from the visibility calibration experiment. Variances for 

 D and A^ were calculated on basis of the variances of 

 n, fiO), and giO), according to the method of Buckland 

 et al. (2001). The variance in number of sightings, n, 

 was expected to differ by year and stratum because of 

 differences in mean turtle density; however, it was not 

 possible to calculate stratum-specific and year-specific 

 variances empirically because not all transects were 

 replicated in all years. The variance of leatherback 

 turtle detections, therefore, was assumed to follow a 

 Poisson distribution, with vdr(n) = n. The variance of 

 /■(O) was estimated analytically with DISTANCE soft- 



ware, and the variance of g^(O) was estimated from the 

 individual g(0) estimates for the three tagged turtles. 

 Overall abundance estimates for the entire study area 

 were calculated as the sum of the stratum-specific abun- 

 dance estimates for the seven full survey years: 1990, 

 1991, 1993, 1995, 1997, 1999, and 2002. During the 

 three years when surveys were flown only off central 

 California (2000, 2001, and 2003), coastwide abundance 

 of leatherback turtles was estimated as the sum of the 

 abundances for the Monterey Bay and Gulf of the Faral- 

 lones strata, divided by the mean proportion of the total 

 abundance found in these two strata during full survey 

 years. The variance in this proportion was estimated 

 across years (n=7) and incorporated into the variance 

 of A'' and D by using standard formulae. Abundance 

 estimates were examined for trends and potential large- 

 scale environmental influences by linear least squares 

 regression that included year and the 12-month average 

 Northern Oscillation Index (NOI; Schwing et al., 2002) 

 as predictor variables. Regressions were performed for 

 all years, and for the subset of seven full survey years, 

 because there was greater uncertainty in the abundance 

 estimates for 2000. 2001, and 2003. 



Results 



Survey summary 



A total of 31,885 km were surveyed in Beaufort sea states 

 of 0-3 during 1990-2003 (Table 1), and annual totals 

 ranged from 2347 to 5151 km during the full survey years 

 when all strata were surveyed, and from 283 to 1030 

 km during the partial survey years, when only waters 

 between Pt. Sur and Pt. Arena were surveyed. Weather 

 conditions varied by year and were the primary deter- 



