Ensemble data assimilation applications to atmospheric and carbon cycle science

Data assimilation methods are interdisciplinary in nature since problems involving dynamical models and observations are often shared across different scientific disciplines (e.g., atmospheric, oceanic, hydrological, and carbon cycle sciences). Sharing the knowledge and experience from different applications of data assimilation methods is, therefore, of fundamental importance for gaining new knowledge and for further improvements of data assimilation methods.

Toward this goal, this study examines applications of an ensemble based data assimilation method entitled Maximum Likelihood Ensemble Filter (MLEF) in two different areas: atmospheric and carbon transport research. One of the most significant differences between these two applications is in the ways the forecast error covariance (i.e., prior error information) is used. This difference arises from different dynamical models used to propagate forecast error covariance from one data assimilation cycle to another. Additional differences include different special and time scales and different typical number of degrees of freedom. Experimental results employing the MLEF approach to atmospheric and carbon transport models will be examined and discussed.