Ensemble Kalman Filters: The Movie 
Doug Nychka
Institute for Mathematics Applied to Geosciences
National Center for Atmospheric Research
Boulder CO
Monday, December 5, 2005
4:10 p.m.
E203 Engineering Building
ABSTRACT
The Ensemble Kalman filter (EKF) is a Monte Carlo based algorithm for data assimilation. Technically, the EKF is an approximate solution of the basic Bayesian filtering problem for dynamical systems. It has the potential, with suitably tuned algorithms, to facilitate data assimilation and to solve inverse problems with only a modest amount of investment in software development and with few assumptions on the linearity of the system. From a statistical point of view, sequential and local ensemble updates are an intriguing practical approximation to solve a well defined Bayes problem. A goal of this talk is to indicate the simple connection between the EKF update and a more conventional spatial statistics analysis (best linear interpolation) of a geophysical field. This connection helps to illustrate the basic operation of the filter as being equivalent to a simple linear regression and makes it easy to describe potential areas for new research.
