The kill figures given in table 25 are those 

 for males from each year class at ages 2 to 

 5, The number of seals taken varies with the 

 survival success of the year class, with the 

 availability of seals on the hauling grounds, 

 and with whether the harvest is concentrated 

 on the larger (older) or smaller (younger) 

 animals. 



The regression equation derived from these 

 data is: 



Y = 24,000 - 690X (r^ = 0.765 r = 0,875) 



where Y = percentage in harvest of the num- 

 ber of pups alive in autumn, i.e., 

 IOOK/Nq-D referring to equa- 

 tion (3) 

 X = percentage that the dead pups rep- 

 resent of the original number of 



pups, i.e., 100 ^ referring to 



equation (3). 



Before this equation is applied to the 

 forecasting problem, it is useful to review 

 the estimates of the number of pups born 

 (table 26). 



The estimate for 1961 agrees closely with 

 the accumulative estimate given in table 25, 

 but the estimate for 1962 does not. Both 

 estimates, however, were based on shearing 

 and sampling, a method still in a develop- 

 mental stage at that time. 



Application of the regression procedure to 

 estimates of the 1 963 year class (Nq =264,000, 

 D = 34,000) yields an estimated kill of 34,700 

 from the year class. This estimate seems 

 especially low because the kill from this year 

 class has already totaled 26,784. Sampling 

 errors may be associated with the regression, 

 or the estimate of the pup population in 1963 

 may have been lower than the number actually 

 born. 



Application of the procedure to estimates 

 of the size of the 1964 year class (Nq = 285,000, 

 D = 23,000 so X = 11.3) yields a forecast of 

 the total kill of 48,200 from this year class. 

 If 3,000 of this number are 2- and 5-year- 

 olds and 67 percent of the rest are 3-year-olds, 

 we suggest a kill of 30,300 3-year-old males 



Table 26. — Estimates of the total number of pups born, 

 based on shearing and sampling and on the counts of 

 dead pups, year classes 1961-65, St. Paul Island 



in 1967. The standard error of the estimated 

 total kill is 5,500. 



Combined Estimates 



The estimates together with their standard 

 errors are: 



1. From mean temperature regression 



21,200 S.E. 9,700 



2. From pup weight regression 



24,700 S.E. 10,600 



3. From dead-pup count regression 



30,300 S.E. 5,500 



The weighted mean estimate is 27,500. 

 The standard errors used here are those 

 of the total kill of 3- and 4 -year -olds, but 

 since all forecasts are derived from the total 

 kill at these ages by multiplying by the factor 

 0.67, the reciprocals of the standard errors 

 sequence are the correct weights. The mean 

 estimate has a larger standard error, how- 

 ever, because of the variability associated 

 with the proportions taken at age 3 and at 

 age 4 (i.e., the unreliability of the factor 0.67 

 used to convert an estimated total to a fore- 

 cast of the kill of 3-year-old males). 



FORECAST OF THE TOTAL KILL ON 

 THE PRIBILOF ISLANDS 



In 1966, the kill on St. Paul Island was 

 80.2 percent of the total, in almost perfect 

 agreement with long-term trends. This per- 

 centage has fluctuated in 1961-65 from a low 

 of 75.2 to a high of 82.7 percent. The 6 -year 

 average is 79.6 percent. Hence, the extrapola- 

 tion to the total for both islands by multiplying 

 the kill on St. Paul Island by about 1.25 is 

 justified, and is now the most reasonable 

 basis for extrapolation available. 



The forecast of the kill of male seals on 

 the Pribilof Islands for 1967 is given in 

 table 27. The total forecast for 1967 is very 

 close to that for 1966, the most accurate yet 

 obtained (table 28). 



Table 27. --Forecast of the kill of male seals in 1967. by age. 

 Pribilof Islands, Alaska 



29 



