[Wrf-users] Problem with real.exe veritical interpolation into valleys in inversion situations

[gene]

We have been running the WRF here in western Montana now for a few 
years, but have been experience very poor model results in valley 
inversion situations.  We think there are several reasons for this.  One 
major factor are poor model initializations.   It appears that when 
coarse input model data is downscaled to finer mesh grids, whether via 
the wrfsi or WPS and real.exe, the current method is producing very poor 
results in mountain valleys in inversion situations.

Example:   In a valley inversion case study we are doing locally, we are 
doing a nested run with domains at 27/9/3/1 km grid spacing using 32 km 
input data.  In this example we were using no feedback between nests.

The first image:  skewt_0712z_40.27.png  is a sounding at the intial run 
time on the 27 km grid at Missoula (KMSO).   The actual station pressure 
at Missoula at this time was 926 mb.  The model surface at 27 km is 
approx 850 mb.  The actual surface temp / dp at KMSO was -7C/-7C with 
dense fog and CALM winds.

The second image: skewt_0712z_40.9.png is for the same time on the 9 km 
grid.  At this grid spacing the terrain is better resolved, so the model 
surface is about 890 mb.  However, the vertical interpolation scheme is 
resulting in a super-adiabatic layer between the model surface in the 
coarse input grid and the 9 km model surface.  The model surface temp 
vs. the actual temp difference is about 12C.  The surface winds in the 9 
km downscale are also much too strong at the model surface.

The third image: skewt_0800z_40.1.png is for the 1 km grid at 0800Z.  
This is the initial run time for this domain.  The results are the same 
super-adiabatic layer.  The temp at KMSO at this time was -5C.

The fourth image:  skewt_0812z_40.1.png is for the 1 km grid at 0812Z.  
The is the 12 hour forecast.  Given the super-adiabatic layer near the 
surface, the model quickly mixed out any semblance of an inversion.  
Actual conditions at KMSO at this time were -8C/-8C with CALM winds and 
dense fog.

The fifth image:  skewt_0812z_40.27.png is for the 27 km grid at 0812Z, 
or the 24 hour forecast.  The solution, while still much too warm, is 
better then at any of the finer resolution grids.  Possibly because of 
the poor init grids for the other resolutions.

Given the difficulties in downscaling into valleys well below the coarse 
input grids we don't expect the best model inits.   In the current 
scheme there seems to be a very significant problem though.  We get the 
same results no matter what vertical grid spacing we use in the low 
levels, whether we use log(p) or p vert interpolation or higher 
lagrangian interpolation.

We have tried doing a true one way nest from 27 km to 9 km, using the 27 
km output to init the 9 km run, but run into the same issue.

Any insights into this problem would be helfpul.

Gene Petrescu
SOO NWS Missoula

-------------- next part --------------
A non-text attachment was scrubbed...
Name: skewt_0712z_40.27.png
Type: image/png
Size: 35161 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20070308/ea0aa669/skewt_0712z_40.27.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: skewt_0712z_40.9.png
Type: image/png
Size: 35617 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20070308/ea0aa669/skewt_0712z_40.9.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: skewt_0800z_40.1.png
Type: image/png
Size: 35584 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20070308/ea0aa669/skewt_0800z_40.1.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: skewt_0812z_40.1.png
Type: image/png
Size: 35543 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20070308/ea0aa669/skewt_0812z_40.1.png
-------------- next part --------------
A non-text attachment was scrubbed...
Name: skewt_0812z_40.27.png
Type: image/png
Size: 35584 bytes
Desc: not available
Url : http://mailman.ucar.edu/pipermail/wrf-users/attachments/20070308/ea0aa669/skewt_0812z_40.27.png