Population growth of Antarctic fur seals: limitation by a top predator, the leopard seal?

Ecology, Dec, 1998 by Peter L. Boveng, Lisa M. Hiruki, Michael K. Schwartz, John L. Bengtson

We could not fit the model described above to data from 1986/1987 through 1989/1990, because pup counts were initiated too late in those breeding seasons to allow reliable estimation of parameters of birth chronology and total numbers of births. Pup counts were sufficient in 1989/1990, however, to apply an approximate form of the model. To estimate the daily rate of pup mortality at NA in that year, we calculated the slope of a regression of the logarithm of NA pup counts (after 1 January, when nearly all pups had been born) as a function of date. To estimate the daily numbers of pups that would have been observed at NC in 1989/1990 if the mortality rate had been the same as at NA, we first smoothed the NC pup counts by robust locally weighted regression (Cleveland 1979). We then used the mortality rate from NA to extrapolate from the peak of the smoothed curve and compared the extrapolated pup count to the actual pup count at the end of the season, as described above, to estimate the number of pup deaths from leopard seal predation at NC in 1989/1990.

Because of serial correlation in residuals and structural correlations between model parameters, the estimates of variance typically reported by regression programs were unlikely to be reliable for our model. Therefore, we calculated precision of the model estimates by Monte Carlo simulations. Daily numbers of live pups were simulated by a stochastic version of the model, in which the probabilities of birth and death on a particular date were determined by the average estimates (across years) of mean birth date, standard deviation of birth dates, daily mortality rate, and total births. Normally distributed "counting errors" were added to these pup numbers to simulate pup counts; the variance of the counting errors was chosen such that the variance in simulated model residuals would match the variance in the observed model residuals. For each colony, NA and NC, 1000 series of daily pup counts were simulated and the regression model described above was fit to each series. The standard deviation of the 1000 simulated estimates of each parameter was used to represent the approximate standard error for that parameter.

Fur seal abundance and trends in the Elephant Island area

In addition to the counts at colonies on Seal Island, fur seals at other colonies in the Seal Islands archipelago and at Elephant Island were counted when opportunities were available, such as during vessel operations to support the Seal Island field camp. Observers used inflatable boats to land near these colonies and visually counted all pups from vantage points or by walking slowly along the periphery of the colony. These counts were typically obtained by one or two observers; on nine occasions when pups were counted by two independent observers, the coefficient of variation between observers was 6%. Fur seal pups on the islet NNW of Seal Island, informally named Large Leap Island, were counted at least once per year from 1986/1987 to 1994/1995, usually in mid- to late January. Two smaller colonies in the Seal Islands archipelago were counted in 1993/1994 and 1994/1995. Antarctic fur seal pups in all known breeding colonies (Bengtson et al. 1990) in the Elephant Island vicinity were counted in 1991/1992 and 1993/1994.