YORK and KOZLOFF: NORTHERN FUR SEAL PUPS 



rookery; and, one small, one medium, and two large 

 rookeries. 



A subsampling estimation procedure was devel- 

 oped for 1980-84: rookeries were grouped into three 

 strata— large, medium, and small rookeries; one 

 small, one medium, and two large rookeries were 

 sampled each year. Furthermore, in order that some 

 rookeries were not disturbed inordinately more than 

 others, each rookery was sampled at least once, but 

 no more than twice during the 5-yr period. We had 

 intended to census all rookeries in 1985, but logistic 

 difficulties permitted a sampling of only seven 

 rookeries. 



A summary of data collected during 1980-86 with 

 the ordinary ratio, jackknife ratio, and regression 

 estimates of the ratio of pups to breeding males ap- 

 pears in Table 2. The estimates based on the three 

 methods are approximately equal within each year; 

 in most cases, the jackknife estimate lies between 

 the ordinary ratio and regression estimates. Esti- 

 mates of the total number of pups born were ob- 

 tained by adding counts of dead pups to number of 

 pups alive at the time of census (based on jackknife 

 ratios); approximate 95% confidence intervals were 

 calculated (Table 2). 



In Figure 4, estimated 95% confidence intervals 



Table 2— Summary of the total number of breeding northern fur seal males, ratios of the number 

 of pups alive at the time of sampling to the number of breeding males counted, estimated number 

 of pups alive at the time of sampling, counted number of dead pups, and estimated number of 

 pups born, and approximate 95% confidence interval based on the jackknife standard errors, 

 St. Paul Island, 1980-84. 



■a 



3 



o 



Q. 



3 



a. 



350 r 



305 



260 



215 



170 



125 



1 



I' 



1967 



1971 



1975 



1979 



1983 



1987 



Year 



FiGt^iRE 4.— Approximate 95% confidence intervals and estimates of numbers 

 of northern fur seal pups born on St. Paul Island, AK, 1970-86. (We include 

 only those years for which data were available to compute estimates accord- 

 ing to the methods developed in this paper.) 



373 



