Development of a Hierarchical Model to Estimate Sea Turtle Rookery Contributions to Mixed Stocks in Foraging Habitats

• develop a hierarchical model to incorporate genetic markers and ecological parameters to estimate rookery contributions to a mixed stock, with appropriate estimates of uncertainty.
• develop and assess methods for selecting or combining hierarchical models with different levels of complexity, including aggregation of genetic markers and/or rookeries.
  
• develop ecological-genetic models that link turtle straying behavior and mtDNA genetics to the observed spatial distributions of mtDNA haplotypes.

Project researchers have recently explored the implementation of a new set of methods, called Markov Chain Monte Carlo estimation (MCMC), with the primary goal of improving confidence limits on existing methods for stock analysis (Bolker et al., unpub.). More recently, project researchers have broadened their exploration, combining MCMC estimation with hierarchical models to combine information on genetic markers with ecological data such as rookery size and location in order to extract as much information as possible from the available data. This class of models provides enormous power to address basic questions in stock analysis and modeling, improve their estimates of turtle origins, and provide more correct estimations of uncertainties. Project researchers will describe their basic approach (MCMC estimation utilizing the Gibbs sampling method) and how it improves estimates of uncertainty over traditional methods; describe extending this approach to include ecological covariates; describe methods for model selection to allow data aggregation where appropriate and keep model complexity in check; and discuss how models with spatial and ecological covariates can be used to make inferences about the founding and evolution of turtle subpopulations.

During the first year of the project, researchers will work with hierarchical models incorporating genetic, ecological and spatial information, exploring a range of options for structural and functional forms of the dependence on size (categorical vs. linear vs. power-law) and spatial aspects of the model (categorical vs. conditional autoregressive vs. distance-based). Project researchers will develop model selection and model averaging methods to guide the development of the models.

The second year of the project will involve applying model selection methods to questions of spatial and genetic aggregation. They will focus on the questions of ecological and evolutionary dynamics by building and parameterizing spatial ecological/evolutionary models using spatial coalescent methods, exploring the accuracy and robustness of the estimation procedures by building ecological/evolutionary simulation models.

Year 1 funding for this 2-year project to be awarded January 2003.

Literature cited:
•Bolker, B., T. Okuyama, K. Bjorndal, and A. Bolten, unpub. Stock estimation for sea turtle populations using genetic markers: Accounting for sampling error of rare genotypes. Submitted to Ecological Applications. Preprint available at http://www.zoo.ufl.edu/bolker/turtle/turtle001.pdf

Principal Investigators:

Dr. Benjamin Bolker
University of Florida
Zoology Department
P.O. Box 118525
Gainesville, FL 32611-8525
USA
Phone (352) 392-5697
FAX (352) 392-3704
email: bolker@zoo.ufl.edu

Dr. Karen Bjorndal
Archie Carr Center for Sea Turtle Research
University of Florida
Zoology Department
P.O. Box 118525
Gainesville, FL 32611-8525
USA
Phone (352) 392-5194
FAX (352) 392-9166
email: kab@zoo.ufl.edu

This page updated August 22, 2006