USING THE ETA MODEL

                        Geoff Manikin 
             NOAA/NCEP/EMC/  Camp Springs, MD and
                  SAIC/GSC    Beltsville, MD

                  Geoff DiMego and Tom Black
              NOAA/NCEP/EMC    Camp Springs, MD

     This presentation was made at the Ninth Conference on
Aviation, Range, and Aerospace Meteorology in Orlando, FL
on 9/13/00.   It demonstrates current output from NCEP's
Eta model relevant to aviation forecasting efforts and 
discusses some upcoming changes to the model.
     Fig. 1 shows the Eta output of interest to
aviation forecasters in both gridded and bufr format. 

     Fig. 2 shows a sample Eta visibility forecast with
the accompanying observations for verification.   The Eta uses
the Stoelinga-Warner algorithm;  units are km.   The areas of
reduced visibility are generally forecast well, but the model
tends too predict values that are too low. 

     Fig. 3 shows an Eta forecast of 10-meter wind speeds
(left) and a forecast of wind gust potential (right).  
Units are knots.   The wind gust algorithm determines the depth
of the boundary layer, finds the maximum wind in the layer, and
mixes a height-weighted fraction of that speed to the surface.  
It is being tested for future addition to the model.

     Fig. 4 shows the observed surface winds (left)
and wind gusts (right) for comparison with Fig. 3.   This
forecast seems to do fairly well just east of the Rockies and
over the intermountain west.   It is noticeably poor over
interior southern California.

     Fig. 5  shows a series of Eta hourly forecasts of turbulent
kinetic energy (TKE) and the accompanying forecasts of 250 mb
isotachs.   The 24-hr forecast shows the development
of a TKE feature.   At 25 hours, the feature is
being advected to the northeast.  At 26 hours, the
main feature has moved well to the northeast and weakened. 
At 27 hours (not shown), the feature is completely gone.
An explanation of ETA TKE is found here.

     Fig. 6 shows the detail that is lost in looking
at a high resolution model forecast on a low resolution grid.
It compares the 2-meter temperature forecasts from the SAME
MODEL RUN on a 22-km output grid (#221) and a 90-km output
grid (#104). 

     Fig. 7 shows the division of the Eta grid 221
files into 'tiles.'   Users who wish to utilize the high resolution
grid 221 output but are unable to deal with the huge files
can download just the region(s) they need.

      Fig. 8 was a picture of the web page which displays Eta
forecast meteograms.   The page can be found here.

      Fig. 9 shows a sample of an Eta forecast of
meteograms for surface variables.   These plots, derived from hourly
BUFR data, are more effective than gridded data for time precision
for events such as frontal passages, precipitation onset/end, and
precipitation type changes. 

      Fig. 10 shows a sample of an upper-air Eta forecast
from the meteogram web site.   Winds, cloud water, cloud ice, and
the 0 and -15 degree C isotherms are displayed. 

      Fig. 11 shows the future plans for the Eta Data
Assimilation System (EDAS).

      Fig. 12 shows the improvements to the analysis that some changes
to the 3DVAR initialization are having.   The top plot compares 
observed and operational Eta analyzed soundings for Nashville, TN.
The bottom shows the same for the new version of the Eta, and a
much better fit to the data is seen.

      Fig. 13 shows the future plans for the Eta
model.   Note that the 22-km Eta has been implemented on 9/26/00.  
Details can be found in the TPB. Also, the extension of the model 
to 84-hours will be delayed, probably until early 2001.

      Fig. 14 shows the domain for the 22-km Eta.
The domain, which had been slightly reduced when the 32-km Eta was
implemented on the IBM supercomputer, is enlarged.

      Fig. 15 compares the Eta model terrain over the
western U.S. in the versions of the model run with horizontal 
resolution of 32 km and 22km.

      Fig. 16 shows differences in forecasts of a cyclone near
Alaska in the 22 and 32 km versions of the Eta.   The differences 
are likely due to increased resolution and the use of radiances.
Forecasters indicated that the 22-km Eta's forecasts of a deeper
storm was better than the prediction by the 32-km model. 

      Fig. 17 shows an orographic precipitation case over
southern California in the winter of 1998 and the huge impact 
resolution has in a model's ability to predict extreme amounts.
The 10-km Eta provided tremendously useful guidance forecasters
for this event.