c. a regression of the percentage kill of 

 males at age 3 against median and 

 terminal dates of the kill of 3 -year- 

 old males. 



Method 3. Forecast of the kill from a year 

 class at ages 3 and 4 based on: 



a. a dead-pup count regression; 



b. an annual mean air temperature; 



c. the estimated number of yearlings, based 



on tag recoveries at age 2. 



Method 4. Forecast of the kill from a year 

 class at age 3 based on: 



a. the kill of males at age 2 or some index 



thereof; 



b. a regression against the mean weight of 



untagged pups. 



When the total return of a year class is 

 forecast (method 1), it is necessary to de- 

 termine the proportion of the year class that 

 will escape the kill and be recruited into the 

 stock of breeding males. We have discarded 

 this method because we cannot estimate the 

 escapement of males accurately, and because: 



1. We question estimates of the size of the 

 pup population based on tag recoveries, par- 

 ticularly those made for year classes 1953 

 through 1959. 



2. Estimates based on tag recoveries at 

 age 2, which provided the only basis for using 

 this method to forecast the kill of 3-year-old 

 males, are extremely variable. 



3. The validity of the population-return 

 equation for recent years is questionable be- 

 cause of probable bias in the population esti- 

 mates derived from tag recoveries from the 

 1953 to 1959 year classes. 



No problem exists when the forecast yields 

 the total kill at ages 3 and 4, and the kill at 

 age 4 is to be forecasted after the kill at age 3 

 has been achieved. The kill at age 3 can be 

 estimated as 67 percent of the total kill from 

 the year class at ages 3 and 4 (an average of 

 67 percent of the males killed on St. Paul 

 Island from yearclasses 1947-62were 3-year- 

 olds). Although actual percentages have varied 

 from a low of 49 (1952 year class) to a high 

 of 81 (1953 and 1954 year classes), much of 

 the variation has been caused by changes in 

 the terminal date of the kill, and, to a lesser 

 extent, changes in the nnaximum body length 

 limit of seals taken. In recent years, both of 

 these factors have been stabilized and the 

 percentage taken at age 3 has been corre- 

 spondingly more stable. 



Method 2a was discarded when we found that 

 estimates of total escapement were too low 

 and when postponement of the start of the 

 season had decreased the reliability of the 



estimates of the postseason escapement. Esti- 

 mates of through-the -season escapennent of 

 undersized animals has been unsatisfactory 

 for some time. 



Method 3a has also been eliminated from 

 current use. A study of methods of forecasting 

 in 1963 (Roppel, Johnson, and Chapman, 1965) 

 showed that data on annual mean tennperatures 

 provide forecasts that are as precise as those 

 based on data from the counts of dead pups. 

 Moreover, the counts of dead pups are definitely 

 weak as a forecasting tool because, as the total 

 population declines, the count of dead pups 

 also declines regardless of the condition of 

 the year class. 



Method 4a was never satisfactory because 

 there has been alnnost no relation between 

 the kill at age 2 and the kill from the year 

 class in subsequent years, nor has it been 

 possible to improve on this approach by using 

 some index to measure the lateness of the 

 returns, such as the number taken during the 

 last 5 days of the kill. 



To achieve the best forecast of the kill of 

 males, a reasonably accurate estimate of the 

 number of pups alive in autumn of the year of 

 birth is necessary, together with an estimate 

 of mortality at sea from 3 months to 3 years 

 of age. A recently developed method of esti- 

 mating the pup population from shearing and 

 sampling for marked-to-unmarked ratios pro- 

 vides reasonably accurate estimates of the pup 

 population. Since the causes of loss at sea 

 remain unknown, estimates of ocean mortality 

 are difficult to obtain. 



Assuming that the variables of annual mean 

 air temperatures and mean weight of pups in 

 autunnn are related to causes of ocean mortality 

 or to viability of young seals, we believe re- 

 gressions based on data fromi these variables 

 are indirect measures of ocean mortality. 



If, however, most of the variation in mortality 

 of pups at sea occurs during the first year of 

 life, and if the total number of yearlings can be 

 satisfactorily estimated from the tagging of 

 yearlings and subsequent tag recoveries at 

 age 2, then a satisfactory forecast can be 

 obtained sufficiently in advance of the kill at 

 age 3 to be useful. We have not yet accumulated 

 sufficient experience with these estimates to 

 rely on them alone. Thus, several methods 

 are now producing several estimates. The final 

 estimate is a weighted average of different 

 estimates with the weights being those sug- 

 gested by statistical theory, i.e., inversely 

 proportional to the variances of the estimates. 



FORECAST OF THE KILL OF 4- YEAR-OLD 

 MALES ON ST. PAUL ISLAND 



Three methods currently used to forecast 

 the kill of 4-year-old males are: (1) Regres- 

 sion of the kill of 4-year-old male seals on the 

 kill of 3-year-old male seals and the mean 



25 



