Departmental Seminar

Bayesian hierarchical models have become invaluable in marine science.  They allow us to accommodate uncertainty about data collection and our understanding of ecological mechanisms.  The adoption of Bayesian methods has led to an increase in model complexity to answer tough questions using large and varied marine data sets.  Complex models can be slow to fit, but a solution that accelerates Bayesian computing exists.  Bayesian models can be rearranged to facilitate computing in stages.  We refer to this procedure as recursive Bayes and it can save us time by leveraging the parallel architecture of modern computers.  I will explain the fundamental concept that allows us to organize Bayesian models in a way that aids multistage computing.  Then I will demonstrate how the procedure helps in marine science studies.  These include: 1) tracking green sea turtles along the Florida Gulf coast, 2) using in situ data from vessels, Argo floats, buoys, and drifters to predict oceanographic quantities like sea surface temperature across the Gulf of Mexico, 3) optimizing aerial surveys to study the recolonization of sea otters in Glacier Bay, Alaska, 4) accelerating capture-recapture models to estimate bottlenose dolphin abundance in Galveston Bay, Texas, and 5) predicting the species distribution of harbor seals in Johns Hopkins Inlet, Alaska.   

Biography: