Data Assimilation for Non-Gaussian Systems

Problem

Data assimilation (the update of model outputs with measurement errors to compensate for imperfect models and to ensure that the results of models match data and observations) is being increasingly used in atmospheric and oceanic sciences.  Reanalyses, the results of GCMs and RCMs after data assimilation methods are implemented, are often used when considering future projections for changes in our climate system.  The Ensemble Kalman Filter is a very common method to do so, and I am interested in improving these methods to make them more tractable for nonlinear, non-Gaussian systems.  Furthermore, I am interested in exploring the use of the Ensemble Kalman Filter and Smoother to do initial condition estimation and parameter estimation as well.

application to ice streams

These time-dependent data assimilation methods have been used in the atmospheric sciences extensively, and I am interested in determining their use for glaciology. Current inverse methods used in glaciology are somewhat clunky and enable only static assimilation of data to make estimates of parameters and states of ice flow models. These time-dependent data assimilation methods could be used to enable dynamic updating of parameters or states of ice flow models.