Split or combine? Effects of repeated sampling and data pooling on the estimation of colony numbers obtained from drone genotyping

Male honey bees trapped at Drone Congregation Areas (DCAs) can be used to infer the number of colonies from which drones were derived, and thereby colony density in the environment. Crucial to the accuracy of this method is precise grouping of males into brother groups based on genetic markers, and a sample size that is sufficient so that all colonies in the area are included in the sample. The optimal sample size is a trade-off between cost and accuracy and cannot be known prior to sampling. Therefore, follow-up surveys may be necessary if the data indicate that the first sample was too small. However, the effect of multiple sampling on allele frequency estimates and the accuracy of the method is poorly understood. Here we trapped drones from two independent DCAs every month over 2 and 2.5 years. We analysed our data using the sibship grouping programme COLONY in three ways: (i) using data from the entire year, and counting the number of colonies identified in each month; (ii) using monthly data with allele frequencies from the entire season; and (iii) using data from each month separately. Although there were significant changes in allele frequencies over the year, these changes had no material effect on classifications of drones into families. Therefore, multiple samples can provide more robust estimates of family groupings due to the larger sample size and can be used with confidence where required.