Updating SGS cloud interstitial to add Chab-Becht cloud fraction and …

[joeolson42]

This modification adds:

1. A new cloud fraction (Chaboureau and Bechtold 2005) for the GF convection scheme
2. The capability to separate frozen subgrid clouds into both snow and ice (no longer just ice).

[climbfuji]

On thing to be aware of: ud_mf is in the GFS_interstitial DDT and as far as I can see it gets set/reset in the physics group. Since it is intent(in) here, you probably get invalid data passed in (from a different block)

[joeolson42]

You are wise to be concerned, but in this case, I think it's OK. Here is a brief printout of ud_mf, which has reasonable values of mass-flux (although sometime zero - maybe the convection isn't reaching that high at some i points):

k= 20 ud_mf= 0.121096490473802
k= 20 ud_mf= 0.000000000000000E+000
k= 20 ud_mf= 0.719252198946091
k= 20 ud_mf= 1.83385869435230
k= 20 ud_mf= 1.69616072660016
k= 32 ud_mf= 1.70287760640253
k= 32 ud_mf= 1.25508844531819
k= 32 ud_mf= 1.30862285882288
k= 32 ud_mf= 1.79306649029620
k= 32 ud_mf= 0.858321688490849
k= 32 ud_mf= 0.471099592930162
k= 32 ud_mf= 0.901912835814989
k= 32 ud_mf= 1.15600537456719
k= 20 ud_mf= 1.39565460657562
k= 20 ud_mf= 1.32165535435073
k= 21 ud_mf= 0.000000000000000E+000
k= 21 ud_mf= 0.000000000000000E+000
k= 21 ud_mf= 0.000000000000000E+000
k= 21 ud_mf= 0.000000000000000E+000
k= 21 ud_mf= 0.000000000000000E+000
k= 21 ud_mf= 0.689175993672929
k= 21 ud_mf= 0.388614361705108

And here is an example of the cloud fractions at the upper-levels produced by these modifications (right) compared to when leveraging the MYNN stratus cloud fraction (left):

These results are within the bounds of expectations to me.

[joeolson42]

Just to be clear, I agree with Dom that this shouldn't work. I am investigating why it may OR how I can make sure it works.

[joeolson42]

OK... I was puzzled as to how this code seems to work, so I:

1. added a "ud_mf_gf" in the TBD block,
2. set ud_mf_gf(i,k) = ud_mf(i,k) in the GF driver,
3. piped ud_mf_gf into module_SGSCloud_RadPre.f90 instead of ud_mf.
4. Ended up with almost identical results

I would happily attach an image for proof but this web interface is not allowing me to do so (error: "Something went really wrong, and we can't process that file).

So, even though either method produces basically the same answer (although not bit-4-bit identical), I suggest we go with this updated method just because we know the TBD block is saved in memory. Note that I failed in converting ud_mf into a TBD array. It failed at compile time and I don't know why. Apparently, it takes more than simply modifying GFS_typedefs.F90 and GFS_debug.F90...

Happy to hear any comments on this.

[climbfuji]

You'll have to move it in GFS_typedefs.F90, GFS_typedefs.meta, and in GFS_debug.F90. In the metadata file, the entry ud_mf must be moved into the metadata section for the tbd type.

[joeolson42]

Thanks Dom. I missed the ud_mf move in GFS_typedefs.meta. Now it compiles and runs and I don't need the additional array. As far as I can tell, this code is ready for the regression tests.

[joeolson42]

Sam,
Perhaps we can ask Haiqin and Hannah to review this PR?
-joe

[hannahcbarnes]

Just curious, did moving ud_mf into Tbd have an impact on results?

[joeolson42]

No, by eye the cloud fractions looked identical, but it was not bit-4-bit identical.

[hannahcbarnes]

I noticed that min(1., max(0., (T3D(i,k)-244.)/25.)) switched to min(1., max(0., (Tk-244.)/29.)). Is the switch from 25 to 29 in the dominator done purposely, or is it a typo?

[joeolson42]

It was done on purpose, but I have plenty of doubt as to which is optimal. In reality, the partitioning of water vs frozen condensate is more than a function of temperature, so both values are wrong.

In this case, since most condensate between 0 and -4 C is in water phase, maybe we should change it back to 25. If you prefer it changed to 25, I 'll do it.

This tweak won't require new regression tests.

[hannahcbarnes]

I am fine with this change, just wanted to make sure something didn't slip through unknowingly

[hannahcbarnes]

This looks good to me. I am looking forward to testing it out on our GFS+GF runs.

[haiqinli]

It looks good to me. Thanks.