Method 1 .-- Temperature -return regression. 

 This forecast equation, which was given on 

 page 27, implies a return of 36,800 for the 

 1962 year class (for which T = 21). This 

 return, or adjusted kill, is the kill of 3- and 

 4-year-old males to 3 1 July, plus 80 percent 

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

 Where the termination date has been 5 August, 

 the total 3- and 4-year-old male kill has 

 exceeded this adjusted kill by about 8 percent. 



Hence, this suggests a kill from the 1962 

 year class at ages 3 and 4 of 39,700. In years 

 with a 5 August termination date, the 3-year- 

 old male kill is about 61 percent of the total 

 3- and 4-year-old male kill. This leads to an 

 estimated St. Paul Island 3-year-old male kill 

 in 1965 of 24,000. 



Method 2 . --Yearling estimate. In 1963, 551 

 yearling males were tagged or retagged. In 

 1964, 56 recoveries were made of thesemales 

 that were of the correct age, that is, 2 years 

 old. In addition, there were three of age 3, 

 that is, animals that were age 2 at the time 

 of tagging. The proportion older than 1 at the 

 time of tagging is similar to the situation in 

 the previous year, that is, from the 1962 year- 

 ling tagging 44 2-year-old males and 2 3- 

 year-old males were recaptured in 1963. The 

 subsequent data have shown that the propor- 

 tion older than 1 in 1962 was 3.2 percent. 

 Using this adjustment, the estimated size of 

 the yearling class in 1963 was: 



... 3,679(534) .. cnn 

 N 1963= - 1 — 57 L = 34,500 



Here 3,679 = 3,678+1, where 3,678 is the 

 1964 2-year-old male kill. This is obviously 

 a low estimate. The estimate made on the 

 basis of similar data in 1963 of the size of 

 the yearling class in 1962 was similarly low. 

 It was in fact: 



N 



1962 



(2,020)(606) 



45 



27,200 



This suggests that the 1963 yearling group 

 may have been 27 percent larger than the 

 1962 yearling group ( 34.5 = ^ 27) 



27.2 ' 

 With other mortality factors remaining about 

 the same, this implies a St. Paul Island 1965 

 3-year-old male kill of (24,500)(1.27) or 28,600. 



Other methods . --A complete kill of all 

 available 2-year-old males in late July would 

 seem to be a possible index of the later har- 

 vest to be expected from a year class. To 

 obtain uniform data over a period of years, 

 a regression of the 3-year-old male kill to 

 31 July was calculated on the kill of 2-year- 

 old males in the last round (27-31) of July 



of the previous year, that is, from the same 

 year class. The regressionis Y = 14.9 + 1.713X, 

 which is based on data from the 1952-61 year 

 classes. Here Y is the 3-year-old male kill 

 (in thousands) to 31 July, and X is the kill 

 (in hundreds) of the 2-year-old males in the 

 last round (27-31) of July of the previous year. 

 For the 1962 year class, X= 4.3, hence Y is 

 predicted to be 21.9. Adding 10 percent forthe 

 expected kill in August yields a season total 

 of 24,100. This is in excellent agreement with 

 estimates found by methods 1 and 2. The value 

 of r<- for this relationship, however, is only 

 0.20, that is, only 20 percent of the variation 

 of Y is associated with or due to X. Thus the 

 relationship can be expected to yield quite 

 erratic predictions in general. 



As mentioned earlier, an estimate of the 

 predicted 3-year-old male kill for 1964 was 

 based, in 1963, on the ratio of past kills to 

 the cumulative female population estimate. 

 Though this was unsatisfactory, it was thought 

 that the cumulative female estimates might be 

 used with the temperature data to give im- 

 proved estimates. The cumulative estimates 

 taken from a previous study (Chapman, 1964) 

 are given in appendix table 3. 



Appendix table 3.- -Cumulative female population 

 estimates. 1950-60 



A regression of the adjusted kill, as shown 

 in appendix table 2, was calculated using 

 these estimates and temperature data (also 

 given in appendix table 2). The resulting 

 equation is : 



Y = -56.0 + 1.20 T + 0.84F 



where Y = adjusted kill 



T = temperature (in tenths of a degree 

 above 32 ) 



F = estimated number of females 



The addition of the new variable is significant 

 when tested by the usual analysis of variance 

 methods (F = 12.08, which is highly signifi- 

 cant) and for the multiple regression R2 = 0.85, 

 which is very high. When applied to the data 

 for 1961, however, the estimated Y is 17,800, 

 which is lower than the number of 3-year-old 

 males already taken in 1964. Whether this is 

 an aberrant observation or whether the rela- 

 tionship is already changing as the herd re- 

 sponds to the reduction program that has taken 

 place must be left for future observations to 

 decide. 



28 



