NOTES from Global Climate+Weather Modeling Branch Map Briefing - 5/22/08

(most speaker's figures are posted; where user=our_EMC, pw=same as rm209 laptop),click here

Glenn White- GFS scores for operational and parallel systems:

1. Day-6, 500mb height AC for Apr21-May21 shows GFS .032/.005 behind EC/UK in the northern hemisphere, AND .083/.027 behind EC/UK in the southern hemisphere. Half the SH difference in scores between GFS and EC seems due to 6 'dropouts', where GFS is 0.2 or more worse than the EC forecasts.

2. Precipitation (days 1-3): Equitable Scores for April 2007 and 2008 show both the EC and GFS improved relative to other Centers in 2008. It was noted that there was a GFS implementation in May 2007, and a significantly improved EC precipitation parameterization in December 2007. In April 2007, a medley of Center scores was better than the ECMWF, seeming to show that other Centers were able to add information to the EC forecasts, but in April 2008, the EC beat everyone, implying that other Centers could not add anything to the EC forecasts.

3. Several cases of GFS, new GFS, and CPC's gauge total precipitation amounts in April and May 2008 were presented, in an attempt to see what HPC didn't like about the new GFS. Some showed that the operational GFS was better, while others showed that the new GFS was better. Pete Caplan noted that of 25 precipitation cases, the new GFS was better for 16 of them. Heavy rain has always been difficult for the models. He further noted that Mike Davison (HPC) had seen the new GFS overintensify small scale features and be inconsistent in the day-to-day forecasts (i.e. 'flip-floping'). Other precipitation scores showed that day-3 rainfall for Jul-Oct 2005 was worse also!

4. During 1/1/08-5/20/2008 the 500mb height AC for day-6 was 0.1 worse for the new GFS. For the 2005 and 2007 retrospective forecasts, the AC scores were much closer.

5. New GFS vs operational GFS; A SUMMARY: The parallel shows slightly worse precip scores over CONUS during Jan-May 08. The midlat height scores are not significantly different. Tropical wind errors may be reduced in the parallel(?). The hurricane tracks in the parallel GFS are improved. The parallel's stratosphere is also improved.

Min-Jeong Kim- Inclusion of cloudy radiances in the NCEP GDAS:

1. Cloudy microwave observational data currently is not used in the NCEP GDAS. Approximately 30-40% of the microwave radiances are cloud-contaminated, which represent a valuable lost resource. This data contains information about total precipitable water and cloud liquid water path. This report contains the first attempts to use cloudy microwave observation data in the NCEP GDAS.

2. Cloudiness and precipitation indicates that some dynamically important weather is occurring. Subsequent forecasts are often sensitive to initial conditions in these regions.

3. Thin cloudy areas have been assimilated without including a cloudy radiance computation, and thick cloud data are screened out. Can we extract cloud information from observations by performing cloudy radiance assimilation? Need to be aware that the correct radiance is a function of cloud particle shape!

4. Cloud profiles in the 1st Guess are determined using the solar radiation parameterization in the forecast model (see Hou, etal 2002; NCEP Office Note #441). Profiles of moisture, ice cloud and liquid water cloud are also shown.

5. Future work:

--------a. Need to employ the GFS model's detailed cloud/precipitation microphysics parameterization in the GDAS.

--------b. Determine which channels to use (23 GHz, 31 GHz, 89 GHz (precipitation)).

--------c. Revisit Bias Correction and Quality Control.

--------d. How to make a link to dynamic variables.

--------e. Run impact experiments.