Temporal changes in allele frequencies provide estimates of population bottleneck size

Monitoring temporal changes in genetic variation can provide estimates of effective population size (N(e)) that are useful for detecting population bottlenecks. We used Monte Carlo computer simulations to quantify the bias and precision of the N(e) estimates obtained from temporal changes in allele frequencies at allozyme and microsatellite loci. The estimates of N(e) overestimated the true N(e) by 12-200% after 1-10 generations of a bottleneck of effective size 4. This magnitude of bias, however, is not likely to cause bottlenecks to go undetected or to cause misguided management recommendations. The bias was nearly negligible for bottlenecks of effective size ≥10, until approximately five bottleneck generations had passed. Three causes of the bias are the loss of alleles, the assumption that the increase in variance in allele frequencies with time is additive (instead of multiplicative), and the assumption that selfing can occur when it cannot. We derived a new equation that substantially reduces the bias. The precision of N(e) estimates was sufficient that, when true N(e) was 4, the 95% confidence interval bracketing a N(e) estimate was <25 in 95% of 500 simulations (when only five microsatellite loci were monitored and when 30 individuals were sampled with a sampling interval of one generation). When the true N(e) was 20, however, only 80% of the confidence intervals were <50 when we sampled 45 individuals and 20 microsatellite loci (i.e., loci with five alleles). High power to detect a bottleneck of effective size 4 can be achieved by monitoring only five microsatellite loci having five alleles each and sampling 30 individuals with a sampling interval of one generation. But achieving high power to detect bottlenecks of size N(e) = 20 requires approximately 20 polymorphic microsatellite loci and 45 individuals with a one-generation sampling interval.