GFDI Student Seminar: Leah Rumanick
Ocean Circulation via Bayesian Inversion and Markov Chain Monte Carlo
Abstract: Recently, statistical models have become more prevalent in physical problems for their ability to incorporate a quantitative measure of uncertainty and to minimize the preprocessing and interpolation stage commonly associated with popular techniques. This presentation introduces how to construct a statistical circulation model around physical oceanography constraints. We start with basic physical properties and known tracer values to set up and parameterize an advection-diffusion equation. Unknown parameters in the equation are determined using a Bayesian inversion framework with an incorporated Markov chain Monte Carlo simulation. Here, we focus on finding the flow of a neutral density layer in the South Atlantic and demonstrate the potential of statistical models as a tool for a variety of oceanography problems.