AN EXAMINATION OF THE VARIABILITY OF MIGRATORY TIMING STATISTICS ESTIMATED FROM CATCH AND EFFORT OBSERVATIONS (SALMON)

The estimate of the mean arrival time based on catch or CPUE of fishes migrating into a fixed harvest area is a function of the number of days fished. Simulation studies using chinook salmon catch and effort data from the Yukon River delta, Alaska indicate that fishing effort concentrated at the tails of the migratory distribution which would tend to erroneously weigh the estimated mean arrival time in the direction of the sample, away from the true mean, is only a significant problem when the number of days open to fishing is small, covering less than 12% of the total duration of the migration. At sampling rates of the time domain greater than 12%, estimated mean arrival times are usually within 50% of the true mean.;Sampling from the average empirical proportion of catch yields a narrower confidence interval on the mean arrival time than does sampling from CPUE data. However, samples from annual daily proportions of CPUE with broader variances yield stronger confidence in arrival time estimates than do samples from migrations of average to small variances. . . . (Author's abstract exceeds stipulated maximum length. Discontinued here with permission of author.) UMI.;The variance of the ratio estimator and the mean square error (biased MSE) for the ratio estimator both allow for the construction of confidence limits for an estimated arrival time based on commercial catch and CPUE data. Arrival time estimates for migrations with large variances and with fewer than 12% of the time domain of the migration sampled have narrower 95% confidence intervals than the same methods produced for arrival time estimates for migrations of small variances. The variance of the ratio estimator is more conservative with sampling rates below 12%, however, it closely matches the biased MSE when sampling greater than 12% of the time domain of the migration. Once about a quarter of the migratory time span is fished, the confidence interval is greatly reduced. This is particularly true for migrations of small variance where the proportions of the population sampled tend to be quite concentrated about the central mass of the time distribution of abundance.