From: Geoffrey Manikin <Geoffrey.Manikin@noaa.gov>
Date: Thursday, October 23, 2003 4:36 pm
ubject: GIF image 896x716 pixels


Regarding the issue of Eta forecasts of high CAPEs in the east,
seeminly corresponding with the 3DVAR or other changes to the 
model, I ran some tests to attempt to isolate the issue.  The 
first test was to add the sfc temperature data back into the 
data assimilation.   

I  ran a September case, for which the model generated erroneously 
high CAPE over the southeastern U.S.
  
http://wwwt.emc.ncep.noaa.gov/mmb/gmanikin/3dvar/comparecape_f54.gif
 
Assimiliating the sfc temperature data had virtually no impact 
on this case.  With all of the problems that the shallow convective 
scheme in the Eta has with regards to sounding structure, I reran 
the case  using the Kain-Fritsch convective scheme.  This scheme has 
generally been shown to produce more realistic sounding structures, 
and I suspected that the shallow convective branch of the BMJ scheme 
in the operationalEta might be responsible for the high CAPE in this case.

http://wwwt.emc.ncep.noaa.gov/mmb/gmanikin/3dvar/comparecape_kf__f54.gif 

Bingo!  The CAPE in northwest GA is reduced from 3000+ to the 
500-1000 range. The KF runs show significantly less CAPE from 
southern  NY all the south to Louisiana. (The KF is noticeably 
higher over the  Gulf and over Florida and Texas.)  It should be 
noted  that neither model generates any precipitation over most of 
the region in question,  so it isn't an issue of the KF scheme 
reducing CAPE by initiating deep convection.  (One reason for 
the differences over the Gulf and  Florida is that the operational 
Eta with the BMJ convective scheme does produceconvective precip 
there which would lower the forecast CAPEs.) A comparison of the 
54-hour forecast soundings for Peachtree  City, GA can be seen at

http://wwwt.emc.ncep.noaa.gov/mmb/gmanikin/3dvar/722150_f54.gif

The BMJ sounding shows a fairly classic shallow convective 
signature, with a "looping" moisture profile between 850 and 750 
mb and a temperature inversion just above.  The KF structure looks 
more realistic.  It is easy to see how the cooling in the 800-600 mb 
layer in the BMJ sounding leads to the very high CAPE.

I therefore conclude that the 3DVAR changes are NOT impacting
longer-range forecasts of sfc temp/dew pt, although I realize 
this is only one case.  Nevertheless, it's safe to say that we 
were  seeing an extreme example of problems with the shallow 
convection in this case.  We've seen examples like this in which 
the CAPE forecast is off by  up to 3000 J/kg due to these problems, 
so I would guess that nothing has "changed" lately, but maybe the 
new physics will interact with  the BMJ scheme to make these extreme
situations more common?   It's definitely something to watch for 
down the line. 
 
Thanks to Eric Rogers for assistance in setting up the model 
reruns,to Brad Ferrier for some helpful discussion, and to Steve 
Weiss for looking at some of the cases and noting some shallow
convective  issues.
                   -Geoff