2-year-olds might serve as a guide to year- 

 class strength. The correlation is only 0.06, 

 however; thus, this variable obviously has no 

 value for forecasting. 



Regression of Return on Tennperature .-- This 

 estimated regression was calculated above as: 



Y = l6.6+ 1.05T 

 where 



Y = the kill of males at ages 3 and 4 

 T = nnean temperature 



For the 1963 year class T = 28 (2.80 above 320), 

 to give Y= 46,000, the kill at ages 3 and 4. At 

 the present time, the best estimate of the 

 number that might be taken at age 3 is 

 (0.67)(46,000) = 31,000. 



The standard error of the forecast Y (kill 

 at ages 3 and 4) is 10,000. In addition, the 

 percentage taken at age 3 varies (standard 

 error of the percentage is 10.17). Because of 

 this variability, a standard deviation of 8,500 

 in the forecast of the kill of 3-year-old males 

 in 1966 is calculated. The count of dead pups 

 was also tried as a variable in this regression. 

 Chapman (Roppel, Johnson, and Chapman, 

 1965) found, however, that the count of dead 

 pups yields essentially the same information 

 as the temperature variable and thus does not 

 provide an additional basis for forecasting. 

 Data for 2 additional years do not change this 

 conclusion. 



Regression of Return on Pup Weights in 

 Autumn . - - Pups were weighed in autumn to 

 measure the relation of weight to survival. The 

 data for this regression are shown in appendix 

 table 3. 



The regression equation is: 



K = 9.7w.66.65 



(r=0.81) 



where 



K = kill in thousands at age 3 



w = mean weight of untagged males 



The rather high correlation is due primarily 

 to one point, that of the 1958 year class. The 

 forecast for the 1963 year class is 19,700 

 (w = 8.9). The standard error is 8,700. In 

 weighting this estimate with the estimate from 

 the temperature-return regression, however. 



I have to allow for the fact that this regression 

 is based on a much shorter series (4 degrees 

 of freedom compared with 10). When this 

 adjustment is made, the weights of the two 

 estimates are 0.60 (temperature) and 0.40 

 (mean pup weights). The weighted average of 

 the two forecasts is 26,400. 



Prediction of Total Kill 



As in previous years, the estimates for 

 St. Paul Island are extrapolated to St. George 

 Island by the long-term proportion of 0.20 for 

 the latter. In 1963 and 1964, the kill of males 

 on St. George Island was in excess of 20 per- 

 cent of the kill on both islands, but in 1965 it 

 fell to 17.3. The cause of variations between 

 the two islands is difficult to determine. The 

 predicted kill of male seals on the Pribilof 

 Islands in 1966 is given in appendix table 4. 



The forecasted and actual kills of males are 

 compared in appendix table 5 for 1961-65. 



Appendix table 3. — Mean weight of untagged 

 male pups and kill of 3 -year-old males from 

 the year class, St. Paul Island, 1957-62 



Appendix table 4. — Predictions of the kill of 

 males in 1966, by age, Pribilof Islands, 

 Alaska 



29 



