in August were added to the kill of 4-year- 

 old males. This has been done on the assump- 

 tion that if seals had not been killed in August, 

 80 percent of the 3-year-olds alive after 31 July- 

 would have been taken as 4-year-olds the fol- 

 lowing year. Also, the starting round of killing 

 before I960 was 27 June to 1 July; the be- 

 ginning round in I960 and thereafter was 2-6 

 July. Therefore the data in I960 and subse- 

 quent years were adjusted by allotting 45 per- 

 cent of the kill in the 2-6 July round to the 

 27 June to 1 July round. The 45 percent factor 

 was based on kills in 1955 through 1959. 



The regression determined from the data 

 is: 



where x 



- 14.83 + 0.56 x x + 5.53 x £ 



= the kill of 3-year-old males to 



31 July 

 = mean date of the 3-year-old male 



kill in days past 20 July 

 = the kill of 4-year-old males to 



31 July, plus 80 percent of the 



3-year-old male kill after 31 



July 



The correlation coefficient, R = 0.82, is 

 highly significant and indicates that 82 percent 

 of the variance of Y is explained by xj and x2. 



For the I960 year class xj = 12, X2 = 4 so 

 that Y = 14.01. 



Since the kill of 3-year-old males in August 

 was 2,714 (of which 80 percent is 2,171), the 

 kill of 4-year-old males by 31 July 1964 is 

 estimated at 11,800. Adding 10 percent for 



4-year-olds that will be taken 1-5 August, the 

 total estimate is 13,000 of this age for St. Paul 

 Island. 



Method 3 .- -In previous years, dead-pup 

 counts have been used to forecast returns of 

 a year class. The apparent relationship of the 

 dead-pup count to the annual mean temperature 

 of the 12 months preceding birth of the pups 

 make it seem worthwhile to consider both of 

 these variables as a means of predicting the 

 kill of 3- and 4-year-old males. The basic 

 data and an analysis of variance to determine 

 if both variables are necessary are shown in 

 appendix tables 2 and 3. 



The small F value (.02) given by the test 

 to determine the effect of neglecting the dead- 

 pup count shows that all the forecast informa- 

 tion is essentially contained in the temperature 

 variable. Consequently, use of temperature 

 only in the regression is preferable, because 

 it provides a datum point for 1952. Using the 

 data for all 10 years, the temperature-kill 

 relationship is: 



Y = 17.4 + 1.01X 



where Y = adjusted kill 



X = 10 (mean temperature in F - 32) 



The mean temperature in I960 was 33.7, so 

 x = 17 and Y = 34.6. The kill in 1963 prior 

 to 31 July was 11,596 and 80 percent of the 

 kill in August (2,714) is 2,171. There remains 

 according to this forecast 34,600 - 11,600 

 - 2,200, that is, 20,800. 



Appendix table 2. — Kill of 3- and 4-year-old males, 

 dead -pup counts, and mean temperature, St. Paul 

 Island, year classes 1950-59 



1 Kill prior to 31 July, plus 80 percent of the num- 

 ber of 3 -year -olds taken in August. 



2 The difference between the annual mean temperature 

 and 32°, multiplied by 10; 10 (Temperature -32). 



38 



