Having more sophisticated models that can resolve processes at even higher resolutions introduces the need to address the nonlinearities of dynamical systems and observation operators. A closed form solution to deal with this problem does not yet exist in nonlinear data assimilation. The DA group at CIRA investigates approaches that can potentially lead to a solution to the nonlinearity problem in data assimilation. Some of the approaches we investigate include the following:

The inevitable “curse of dimensionality” in data assimilation: The rapid growth in the number of earth observing systems, computational power, and data storage capabilities allows for the use of more complex numerical models and increases the potential of assimilating far more observations that previously possible. However, these benefits can also be a handicap as the give rise to the so-called “course of dimensionality,” in other words, the difficulty of dealing with large-dimensional problems. Our group develops techniques to account for complex interactions between multiple spatiotemporal scales and limitations introduced by dealing with high-dimensional dynamical/observation systems.

In order to produce accurate and timely weather forecasts for extreme weather events (e.g. hurricanes, severe thunderstorms, tornadoes), satellite observations must be combined with prior numerical forecasts from models. The DA group at CIRA develops applications in the area of mesoscale analysis and prediction of severe weather events through the use of ensemble, variational, and hybrid data assimilation.

Researchers from the data assimilation groups at CIRA actively participate in training of individuals as part of a National Oceanic and Atmospheric Administration (NOAA) award intended to increase the number of experts in the field of data assimilation.