Prospects and Challenges for Data Assimilation in Magnetosphere Models

Data assimilation (DA), i.e., the ingestion of data into large-scale numerical models was first introduced into numerical weather prediction models in order to overcome the problem of insufficient data for model initialization. It is now a standard procedure that is used in all contemporary forecast models. Following the success in atmospheric models, DA methods have been introduced into oceanic, and most recently into ionospheric specification and forecast models. Data assimilation has also been introduced in some regional magnetosphere models such as radiation belt models; however, no attempts have been made yet to use data assimilation in global coupled magnetosphere models.

Although the prime reason is the lack of sufficient data, there are also other issues. While DA in the before mentioned areas primarily addresses an initialization problem, the magnetosphere is much more a driven system with relatively short memory. The magnetosphere is also characterized by discontinuous field variables and other constraints, such as the requirement that divergence of the magnetic field vanishes. These issues make the application of most of standard DA procedures difficult and would require development efforts that are generally beyond the means of the development teams.

The development of ensemble Kalman filter algorithms (EnKF) makes it possible to introduce advanced sequential DA procedures into magnetosphere models without excessive programming efforts. In this talk we outline a possible approach towards DA in a global magnetosphere model using the OpenGGCM magnetosphere model with the NCAR Data Assimilation Research Testbed (DART) and discuss expected outcomes and possible pitfalls.