An efficient algorithm for assimilation of satellite retrieval profiles of chemical trace gases in the troposphere with the DART ensemble Kalman filter in WRFCHEM

Abstract for oral presentation at the 18th Conference on Integrated Observing and Assimilation Systems for Atmosphere, Oceans, and Land Surface (IOAS-AOLS) at the 94 AMS Annual Meeting, February 2-6, 2014, Atlanta, Georgia

An efficient algorithm for assimilation of satellite retrieval profiles of chemical trace gases in the troposphere with the DART ensemble Kalman filter in WRFCHEM

Arthur P. Mizzi*, Ave F. Arellano+, David P. Edwards*, Jeff Anderson#, and Jerome Barre*

*Atmospheric Chemistry Division National Center for Atmospheric Research Boulder, CO 80307

+Department of Atmospheric Science University of Arizona Tucson, AZ 85721

#Institute for Mathematics Applied to Geosciences National Center for Atmospheric Research Boulder, CO 80307

mizzi@ucar.edu 303-497-8987

Satellite observation systems provide an abundance of atmospheric trace gas profile data that can be used to initialize atmospheric chemistry forecast models. There is a large amount of available data but overlapping vertical sensitivities lead to vertical correlations and redundant information in the retrievals. The associated retrieval error covariance matrix is non-diagonal, limiting the applicability of sequential ensemble Kalman filter (EnKF) data assimilation algorithms. Rotating the retrieval equation based on a singular vector decomposition (SVD) of the retrieval error covariance matrix diagonalizes the covariance but does not address the redundant information issue. The redundant information can be problematic due to the large amount of storage required for archiving satellite data.

In the retrieval process, the averaging kernel is a matrix that relates the retrieval or expected state to the true state. It describes the state space subspace in which the retrieval lies. It is a Jacobian describing the relationship between changes in the retrieval and changes in the true state. This paper presents an algorithm that uses an SVD of the averaging kernel to identify vectors that form an incomplete basis for the associated retrieval space. Those vectors can be used to remove redundant information from the retrieval and obtain a truncated retrieval based on the independent information content of the original retrieval. That data compression step provides large reductions in the retrieval data storage requirements because the number of degrees of freedom of signal for the retrieval (dofs) is generally less than the vertical dimension of the retrieval (n). It also provides large reductions in the data assimilation computational time because the number of observations to be assimilation is reduced in proportion to n-dofs. After data compression we use an SVD to diagonalize the truncated retrieval error covariance matrix and facilitate use of an EnKF.

This algorithm provides a uniform approach for pre-processing retrieval data from different sensors for different trace gases. The pre-processed observations can be used in any retrieval data assimilation systems without code modification.

We apply this algorithm to assimilate Terra-MOPITT CO retrievals with the NCAR Data Assimilation Research Testbed (DART) EnKF in WRF-Chem for June 2008. We demonstrate its utility by comparing its CO forecast verification scores with those from the direct assimilation of retrievals.