HOHN and HAMMOND: POSTNATAL GROWTH OF SPOTTED DOLPHIN 



the franciscana dolphin, Pontoporia blainvillei, 

 (Kasuya and Brownell 1979). 



Methods 



The data used in this analysis were from all fetuses 

 and calves sampled between 1973 and 1981 except 

 for three specimens <68 cm identified as calves and 

 one specimen of 91 cm identified as a fetus {n = 609). 

 These four specimens were judged to have been 

 misidentified. The data were stratified by area, north 

 and south of lat. 1°S, and the northern sample was 

 further stratified by the size of kill in each net set. 

 Powers and Barlow (1979^) have shown that in net 

 sets in which the kill of spotted dolphins from the 

 northern offshore regions was <40 (small-kill sets), 

 about twice as many calves were killed as a propor- 

 tion of the total kill as in sets where the kill was >40 

 (large-kill sets). This would introduce a bias in the 

 estimate of length at birth in the regression pro- 

 cedure because about 90% of all northern specimens 

 were from small-kill sets. The effect would be to 

 underestimate length at birth because the ratio of 

 calves to fetuses was too high in most of the sam- 

 ple Td investigate the extent of the bias, we 

 calculated average length at birth for all northern 

 specimens, for specimens from sets with kill <40 and 

 >40, and for specimens from sets with kill <30 and 

 >30 because the sample size for sets with kill >40 

 was small. The small sample available for southern 

 specimens prevented any further stratification of the 

 data. 



Length at birth was estimated by fitting a logistic 

 model to the percent postnatal at each length inter- 

 val, weighted by the inverse of the binomial variance 

 of each percentage, and estimated by calculating 

 from the fitted curve the length at which 50% of the 

 specimens were predicted to be postnatal. We also 

 investigated linear and asymmetric logistic-type 

 models. 



improve the fit and gave similar results. 



Ikble 1 shows that the estimates of length at birth 

 for northern specimens using the unstratified data 

 (82.0 cm) and using specimens from sets with kill 

 <30 (81.6 cm) or <40 (81.6 cm) are lower than the 

 estimates using specimens from sets with kill >30 

 (84.6 cm) or >40 (85.4 cm), demonstrating that the 

 bias resulting from an overrepresentation of calves 

 in small-kill sets is significant. Furthermore, the 

 estimate for kill >40 is higher than that for kill >30, 

 indicated that the proportion of calves in the sam- 

 ple may still be a function of kill-per-set at this level. 

 Further stratification to investigate whether or not 

 estimates of length at birth continue to rise at higher 

 thresholds of kill-per-set was not possible because 

 of small sample siz& 



The estimate of length at birth for southern 

 specimens is 83.2 cm. No stratification was possible 

 because of the small sample 



Estimates of standard deviations of the estimates 

 of length at birth are not given because, in fitting 

 the logistic model, sums of squares were minimized 

 for differences between observed and predicted per- 

 cent postnatal and it was unclear how to calculate 

 the standard deviations. 



Table 1.— Estimates of length at birth stratified by number of 

 offshore spotted dolphins killed per set. The range of length 

 classes includes the last 0% postnatal length class and the first 

 100% postnatal length class. 



Results 



Figure 2 shows the length-frequency data for 

 northern specimens from sets with kill <40, and the 

 logistic model fitted to the data. Figure 3 is the 

 equivalent for northern specimens from sets with 

 kill >40. Tkble 1 gives the results for all stratifica- 

 tions described above using the logistic model. Using 

 the linear or asymmetric logistic-type models did not 



^Powers, J. E., and J. Barlow. 1979. Biases in the tuna-net 

 sampling of dolphins in the eastern tropical Pacific Doc 

 SOPS/79/31, Status of Porpoise Stocks Workshop, La Jolla, CA, 

 27-31 August 1979. Unpubl. MS. 



Discussion 



Our estimate of length at birth of 82.0 cm from 

 the unstratified northern data is similar to that of 

 Perrin et al. (1976), who estimated length at birth 

 at 82.5 cm, based on a sample of 73 northern 

 specimens (calves and fetuses) grouped into 3 cm in- 

 tervals from 74 to 92 cm. 



Estimates of length at birth in large-kill sets are 

 less biased because of the overrepresentation of 

 calves in small-kill sets. A future larger sample from 

 large-kill sets may allow for additional stratification 

 by kill-per-set, enabling estimates to be calculated 



555 



