I have some good news. I have been discussing with Matthew Phyle, I want to really thank him, that suggested me some change in the NMM code that are very effective.
1) The most effective: apply some averaging of the vertical motion within the vertical advection
in dyn_nmm/module_ADVECTION.F at about line 1030 of the routine and in the subroutine VAD2, change
LOGICAL,SAVE :: TRADITIONAL=.TRUE.
LOGICAL,SAVE :: TRADITIONAL=.FALSE.
By setting TRADITIONAL to false, a horizontally averaged vertical motion is used to advect humidity and cloud water in the vertical.
This change will slow down the precipitation process and make localized storms less intense (but generally doesn't eliminate them completely.
2) Apply some horizontal diffusion.
you can try to modify line 134
SLOPHC=SLOPHT*SQRT(2.)*0.5*18 (double the original value)
SLOPHC defines a maximum allowable slope to the model surface for which lateral diffusion is permitted
By setting to a value (first case) allows diffusivity to be applied at more points in the horizontal
By setting to 99999. will ensure that lateral diffusion is applied everywhere (In my opinion too much)
3) Always in nmm/module_DIFFUSION_NMM.F
you might want to try to allow for some weak background diffusion in points where the topographic slope would generally lead to a complete zeroing of the impact of horizontal diffusion.
There are two blocks within dyn_nmm/module_DIFFUSION_NMM.F
SNE(I,J)=0. ! ->> change to SNE(I,J)=0.1 ??
and shortly below in the code:
SSE(I,J)=0. ! -->> change to SSE(I,J)=0.1 ??
You should try which is best for you
Of course you have to recompile and use new executable.
We are using these changes operationally (http://www.lamma.rete.toscana.it/wrf-web/index.html
) with KF and in this period are performing very well
I don't know about winter time