Handout (447.4 kB)
The accuracy of the retrieval was validated by comparing synthetic GOES imager 6.7 um brightness temperatures generated from the retrieved water vapor profiles using the RTTOV-7 forward model to GOES observations. Simulated GOES TB's were correlated with observed TB's at r=0.90 and an rms error of 2.9 K.
Results from data assimilation model runs with and without AMSU-B are presented. AMSU-B data dries out the middle and upper troposphere, correcting a known model bias. Moisture gradients over the ITCZ and other features are also strengthened, correcting the model's tendency to smooth out features in the water vapor field. Anomaly correlations of 500 mb heights are neutral to positive.
Tropical cyclone forecasts were also used for verification. Overall, forecast errors for position and central pressure were lower for the AMSU-B run. In addition, the AMSU-B run made a larger number of successful storm predictions, i.e. was able to predict storms a few days out that were not predicted in the control run. Transition of AMSU-B retrieval assimilation to the operational version of NAVDAS and NOGAPS at the Navy's Fleet Numerical Meteorology and Oceanography Center (FNMOC) is targeted for the fall of 2004.