Characterizing sex ratios of sea turtle populations: A Bayesian mixture modeling approach applied to juvenile loggerheads (Caretta caretta)

Sex ratio is fundamental to the demography of animal populations. For sea turtles, the operational sex ratio (i.e., that of breeding adults) can be inferred from juvenile sex ratio. In this study, we characterize sex ratio of n = 1401 juvenile loggerhead sea turtles (Caretta caretta) sampled from foraging grounds in North Carolina, USA. For a subset (n = 170), sex was confirmed by laparoscopy, and for the remaining individuals, sex was classified probabilistically using a mixture regression model. The model predicted plasma testosterone concentration as a function of sex, which was treated as a latent variable, and of other potential covariates, namely water temperature and carapace length. Furthermore, it quantified uncertainty in the sex classification of each individual. For the full data set, the predicted sex ratio was 81% female. Using the subset of turtles with known sex and a leave-one-out validation approach, we found the model to have classification accuracy of 94%. We further used that subset to examine how many laparoscopies are sufficient for accurate classification of turtles with unknown sex, in an attempt to provide guidance for other studies. We found diminishing returns as sample size increased, and recommend 100-140 turtles as a sufficient range.