of tags from 3-year-olds, the kill on St. Paul 

 Island from these year classes at ages 3 and 

 4, and the ratio expressed as a percentage, 

 i^'rom table 24 the estimate of the yearling 

 population of the 1965 year class is 94.0 

 thousand, of which 40.2 percent is 37.8 thou- 

 sand. Therefore, an impliedbalance to be taken 

 from the 1965 year class at age 4 is 18.3 

 thousand. The standard error of the forecast is 

 4.8 thousand. 



Combined Estimates of the Kill of 4 -Year- 

 Old Male Seals on St. Paul Island 



The several estimates and their standard 

 errors are: 



Standard 



The analysis in table 27 might suggest that 

 more weight should be given to the estimate 

 derived from the yearling estimate. Because 

 the combined average estimate is identical 

 with this yearling-derived estinnate, however, 

 such increased weighting would have no effect 

 for 1969. 



FORECAST OF THE KILL OF 3-YEAR-OLD 



MALE SEALS 



Two methods used to forecast the kill of 

 3-year-old males are discussed in this section: 

 (1) Two regressions, one based onair temper- 

 atures and another on mean weights of living 

 pups and counts of dead pups, and (2) estimated 

 number of yearling males. 



Regression of Kill at Ages 3 and 4 on Air 

 Temperature, Mean Weights of Living Seal 

 Pups, and Coimts of Dead Seal Pups 



All regressions except the first calculated 

 in the previous section are also useful for 

 making estimates of the kill of 3-year-old 

 males expected in 1969. Estimates of the kill 



of 3-year-old males in 1969 are derived by 

 multiplying the combined estimate of the kill 

 of 3- and 4-year-olds (K) by 0.67 (the ratio 

 of the kill of 3-year-old males to the kill of 

 3- plus 4-year-old males), a factor that has 

 been used for several years. For the 1961-64 

 year classes the ratio was 0.66, almost iden- 

 tical to the long-term ratio. Use of this most 

 recent ratio would alter the forecast only 

 slightly. The regressions and the estimates 

 are: 



1. Temperature regression K = 17. 6^1+0. 98T 

 For the 1966 year class T = 29, K = 46.0 

 thousand with a standard error of 9.5 thou- 

 sand, and for 1969 the kill of males at 

 age 3 is estimated as 30.8 thousand. 

 2. Pup weight and dead pup count regression 

 K = -12.14-0.25D+6.73W 

 For the 1966 year class D = 22.5, W = 9.6, 

 K = 46.8 with a standard error of 11.4 

 thousand, and for 1969 the kill of males at 

 age 3 is estimated at 31.4 thousand. 



Forecast of the Kill at Age 3 Based on an 

 Estimate of the Yearling Male Seal 

 Population 



Table 24 gives 19,285 as the estimated 

 number of yearlings surviving fronn the 1966 

 year class. This estimate of the yearling 

 population is the lowest to date, though all 

 estimates based on recoveries at age 2 have 

 been biased downward. It is still not clear 

 why this bias exists, unless tagged seals are 

 selected during the commercial kill. The ex- 

 tent to which such selection might vary from 

 year to year is unknown, though evidence to 

 date suggests considerable variation. To obtain 

 a valid procedure for forecasting it is neces- 

 sary to adjust the estimate obtained at age 2, 

 which can be done with a direct correction 

 factor derived empirically from the various 

 final estimates of earlier year classes. An 

 alternative indirect but simpler procedure is 

 to calculate a regression of the kill at age 3 

 against the estimate of the yearling population 

 obtained from tag recoveries atage 2. The data 

 are in table 34 and the regression is: 



K3 = 15.3+0.15E (r2 = 0.65) 



for the 1966 year class E = 19.3 and hence 

 K 3 = 18.3 with a standard error of 6.1 (all 

 in thousands). 



Although the yearling popiilation estimate 

 and the resulting estimated kill of 3-year- 

 old males seem low, it is difficult to discard 

 the two entirely because the estimate derived 

 from yearlings has to date been the best for 

 forecasting the kill of 4-year-old males. More- 

 over, the estimate, although low, is reasonable 

 because the major variable is apparently sur- 

 vival during the first 2 years. All other methods 



26 



