14.4
Comparison of algorithms for assimilating satellite partial column retrievals from MOPITT and IASI with WRF-Chem/DART

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner
Thursday, 8 January 2015: 4:15 PM
131AB (Phoenix Convention Center - West and North Buildings)
Arthur P. Mizzi, NCAR, Boulder, CO; and A. F. Arellano, D. P. Edwards, and J. Anderson

Abstract for oral presentation at the 19th Conference on Integrated Observing and Assimilation Systems for Atmosphere, Oceans, and Land Surface (IOAS-AOLS) at the 95th AMS Annual Meeting, January 4-8, 2015, Phoenix, AZ

Comparison of algorithms for assimilating satellite partial column CO retrievals from MOPITT and IASI with WRF-Chem/DART

Arthur P. Mizzi*, Ave F. Arellano+, David P. Edwards*, and Jeffery Anderson#

*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

We have successfully incorporated WRF-Chem (the WRF regional forecast model with online chemistry) into DART (the Data Assimilation Research Testbed) with forward operators to assimilate MOPITT (Measurement of Pollution in the Troposphere) and IASI (Infrared Atmospheric Sounding Interferometer) satellite instrument carbon monoxide (CO) trace gas retrievals. When assimilating trace gas retrievals, one must address issues that are not generally encountered with conventional meteorological observations. For example most sequential ensemble Kalman filter techniques require that the observation error covariance be diagonal. For MOPITT and IASI CO retrievals the error covariance is non-diagonal. There are various ways to achieve diagonalization, for example: (i) discard the non-diagonal elements (set them to zero); (ii) scale the system by the inverse square root of the error covariance; or (iii) rotate the system by the left singular vectors (eigenvectors in this case) of the error covariance. We compared those methods and found that method (iii) – the SVD rotation – provided the greatest forecast skill. As another example due to under sampling errors ensemble Kalman filters require some form of localization. With trace gas retrievals there is a decision on whether to center the vertical localization function at the vertical location of the: (i) retrieval or (ii) the maximum sensitivity in the averaging kernel. We compared those methods and found that method (i) – centering the vertical localization function at the retrieval location provided the greatest forecast skill. We refer to the algorithm using error covariance localization based on the SVD rotation and vertical localization based the retrieval vertical location as the “DA Algorithm.” We applied that algorithm to a 30-day cycling experiment with 6-hr cycling for June 2008 over the continental United States (CONUS) and found that the assimilation of partial column CO retrievals improved the WRF-Chem forecast skill for the CO chemistry variable and the u, v, and T meteorology variables.