13th Conference on Satellite Meteorology and Oceanography


The Impact of AMSU-A Radiance Assimilation in the U.S. Navy's Operational Global Atmospheric Prediction System (NOGAPS)

Nancy L. Baker, NRL, Monterey, CA; and W. F. Campbell

Temperature retrievals from polar-orbiting satellites have been available operationally for twenty-five years. Retrievals are clearly beneficial in the S.H. and the stratosphere, but consistent positive impact on N.H. forecasts has been difficult to demonstrate. Not only is the retrieval problem formally ill-posed, but the errors in retrievals are complex and difficult to characterize. An alternative to retrievals is the direct assimilation of radiance data. In particular, the assimilation of microwave radiance data from the Advanced Microwave Sounding Unit (AMSU-A) has shown clear positive impact in both hemispheres at other major weather centers, and now here at the Naval Research Laboratory (NRL) and Fleet Numerical Meteorology and Oceanography Center (FNMOC). There are a number of requirements that had to be met before that impact could be achieved. First is a modern variational data assimilation capability, coupled with a global numerical weather prediction (NWP) model. The assimilation system requires a reasonable estimate of both the observation and background (short-term forecast) error covariances, while the prediction system must be able to produce a reasonable short-term forecast. The NRL Variational Data Assimilation System (NAVDAS) and the Navy Operational Global Atmospheric Prediction System (NOGAPS) form such a system, and provide a firm foundation for the direct assimilation of radiance data from current and future satellites. Successful direct radiance assimilation also requires a fast radiative transfer model (e.g. RTTOV-7 or OPTRAN) as a forward model, to convert NWP model fields into brightness temperatures (BTs) in observation space. We must thin (or super-ob) the data enough to mitigate spatially correlated error, but not so much as to lose significant information. Strict quality control is crucial, as one bad observation can have a much stronger negative impact than throwing away many good observations. Because we lack good microwave surface emissivity maps, channels that are sensitive to the surface must be screened out over land, ice, and snow. Finally, we must correct for the scan and air-mass based biases remaining in the innovations (observed BTs minus background BTs, computed from NWP fields and the forward model, and interpolated to the observation location). We use a modification of the multilinear regression scheme of Harris & Kelly operationally at FNMOC, and we are actively seeking improvements to the scheme at NRL. Direct radiance assimilation of AMSU-A from the NOAA satellites has been running in near real time since November 2003, and in beta testing at FNMOC since March 2004. Pre-operational testing shows excellent improvement in NOGAPS forecast skill in both hemispheres, compared to a control run using retrievals. The 500 hPa anomaly correlation at 5 days shows a 3-6 hour improvement in the N.H., and an 8-14 hour improvement in the S.H. There have also been fewer forecast “busts” than with the control run using retrievals. Comparisons with the National Centers for Environmental Prediction (NCEP) model over 4 months of testing show that our 500 hPa anomaly correlation scores are virtually identical, and we have closed the gap in the S.H. significantly. Tropical cyclone tracks also show significant improvement over a control run in the May-June 2002 period.

extended abstract  Extended Abstract (1016K)

Poster Session 3, Data Assimilation
Tuesday, 21 September 2004, 9:30 AM-11:00 AM

Next paper

Browse or search entire meeting

AMS Home Page