The U.S. Navy's new three-dimensional variational analysis system NAVDAS became operational at Fleet Numerical Meteorology and Oceanography Center (FNMOC) on October 1, 2003, paving the way for the direct assimilation of NOAA AMSU-A radiances with the Navy Operational Global Atmospheric Prediction System (NOGAPS). AMSU-A radiance assimilation, which became operational at FNMOC on June 9, 2004, leads to significant improvement in forecast skill, as compared with assimilation of NESDIS ATOVS retrievals. The two- to five-day forecast skill at 500 hPa is increased by 3-10 hours in the Northern Hemisphere, and by 12-20 hrs in the Southern Hemisphere, with similar improvements at 1000 hPa. Forecasts with AMSU-A are consistently better, with fewer forecast "busts", fewer synoptic errors and a general strengthening of the circulations in both hemispheres. Overall, NAVDAS analyses and forecasts with AMSU-A exhibit better fit with radiosondes and other observations. Observations from AMSU-B, which are sensitive to the vertical distribution of water vapor in the troposphere, are used to compute 1DVAR humidity retrievals. NAVDAS assimilation of AMSU-B retrievals into NOGAPS dries out the middle and upper troposphere, and strengthens moisture gradients such as the Intertropical Convergence Zone, correcting known model tendencies. Tropical cyclone track and intensity predictions are slightly improved. Transition of AMSU-B retrieval assimilation to operations at FNMOC is targeted for early 2005.
Temperature retrievals from polar-orbiting satellites are clearly beneficial in the Southern Hemisphere and the stratosphere, due to lack of conventional data, but have neutral impact on Northern Hemisphere forecasts. An alternative to retrievals is the direct assimilation of radiance data. The NRL Variational Data Assimilation System (NAVDAS), coupled with the Navy Operational Global Atmospheric Prediction System (NOGAPS) NWP model, constitute a system capable of three-dimensional variational assimilation (3DVar) of radiance data. In particular, the assimilation of microwave radiance data from the Advanced Microwave Sounding Unit (AMSU-A) has shown clear positive impact on 5-day forecasts in both hemispheres. One requirement for successful radiance assimilation is bias correction. Biases are due both to the satellite instrument, and the underlying airmass, resulting from inaccuracies in the fast radiative transfer model that converts NWP fields into simulated radiances. Our approach to airmass bias correction uses multilinear regression of fifteen days of observed minus computed radiances, with selected NWP fields as predictors. Research into hybrid methods, which add the radiances themselves as predictors, is being pursued. Moisture retrievals from AMSU-B can also benefit from bias correction. Preliminary results comparing uncorrected and bias-corrected AMSU-B moisture retrievals are presented. The need for bias correction is universal. Our methodology is robust and general, and should be applicable to current and future satellites.