TITLE
"Theory and Applications of Variational, Sequential and Bayesian
Hierarchical Methods in Data Assimilation"
ABSTRACT
Data assimilation methods that combine observations with dynamical
models of the atmosphere and ocean rely on estimating the conditional
probability distribution function (PDF) of the dynamical system
conditioned on an observed state provided by measurements.
Constructing such a PDF is impractical for most problems with high
dimensions in the state vector and requires simplifications in the
estimations of the conditional PDF. We compare the theories underlying
several data assimilation methods (e.g. variational, ensemble,
particle, bayesian hierarchical) and show how their estimate of
the conditional PDF differs. We also present an example of a
variational data assimilation experiment where the ocean surface
winds associated with the transit of a hurricane are reconstructed
using ocean buoy data.
Emanuele Di Lorenzo
School of Earth and Atmospheric Sciences
Georgia Institute of Technology
Atlanta, GA, 30332-0340