abstract
- Many species of sea turtles spend part of their life cycle gathered in large feeding aggregations that combine individuals from widely separated rookery populations. Biologists have applied methods of mixedâstock analysis to mitochondrial DNA samples from rookeries and mixed populations to estimate the contributions of different rookeries to the mixed stock. These methods are limited by the amount of genetic overlap between rookeries and fail to incorporate ecological covariates such as rookery size and location within major ocean currents that are strongly suspected to affect rookery contributions. A new hierarchical Bayesian model for rookery contributions incorporates these covariates (and potentially others) to draw stronger conclusions from existing data. Applying the model to various simple scenarios shows that, in some cases, it can accurately estimate turtle origins even when turtles come from rookeries with high degrees of genetic overlap. Applying the model to more complex simulations shows that it performs well in a wide range of scenarios. Applying the model to existing data on green turtles (Chelonia mydas) narrows confidence intervals but does not change point estimates significantly. Applying it to loggerhead turtles (Caretta caretta) strengthens the dominance of the large rookery in south Florida, and brings estimates from a small data set on sea turtle strandings into line with those from rookery data. Used appropriately, hierarchical Bayesian methods offer great potential for introducing multiple levels of variation and ecological covariates into ecological models.