The herd density, t)^, for each replicate transect, i, was estimated as 



where 1^ was the transect length and n was the number of herds sighted. To insure that no 

 herds were coimted more than once (during each replicate), each transect was considered 

 a replicate. The herd density, 6^, for each survey block was estimated from R replicate 

 transects by 





and the variance of this estimate was approximated as 



Dolphin density, t)^, was calculated as 



^a'^h 



the product of the mean herd size (fl) and estimated herd density. The variance of t)^ was 

 estimated using Goodman's (1960) method of estimating the variance of a product 



v^r (Z?d) -^s% (4) 2+i5^s% (/f) ==-s% (i/) ^sfe (4) 2 



where n was the number of herds used to estimate the mean. The standard error was 

 estimated as 



The approximate 95% confidence interval, assuming lognormal error, was estimated as 



where 



1.96 ./ln(i*(-^)*) 



C-e V ^ 



106 



