Calling shr_wv_sat_init() from mediator rather than driver?

[uturuncoglu]

The shr_wv_sat_init() call is required to to use shr_wv_sat_qsat_liquid() routine and similar ones to calculate the saturation mixing ratio from the tables but the current implementation assumes that the CMEPS flux calculation will be only used by CESM and its driver. Since we are coupling MPAS and MOM6 outside of the CESM framework (at this point) through ESMX driver, the shr_wv_sat_qsat_liquid() call returns all zero and flux calculation is messed up. As I see those codes are under driver and it makes sense but I think moving those outside of the driver code and init through the med.F90 rather then esm.F90 would make it more generic and it could be also used by UFS WM (I did not check their driver has this init call or not yet) and other projects like ours.

[uturuncoglu]

JFYI, @mvertens @billsacks

Actually I am using little bit old version of CMEPS and that one has no option like

CMEPS/cesm/flux_atmocn/flux_atmocn_Large.F90

Line 217 in a0343bc

if (aofluxes_use_shr_wv_sat) then

. But still it would be nice to move the init to mediator from driver since aoflux calculation is not specific to driver.

[uturuncoglu]

I found a comment in here about the placement of the init call:

CMEPS/cesm/driver/esm.F90

Line 486 in a0343bc

! TODO: this does not seem to belong here - where should it go?

[billsacks]

I don't have enough understanding of this to give an informed opinion. If you have a sense of where this should go, I'd support a change.

[uturuncoglu]

@billsacks As I understand that init call needs to be called in mediator level not in the driver. You could use mediator aoflux phase (incl. also CEMS specific aoflux algorithms) without using CESM driver. In our case we used the flux calculation but it was returning 0 for saturation mixing ration since the table is not initialized. It would be also check in there if the table is not initialized it could return an error to the user. That is also missing. In bigger picture, I think CESM specific aoflux algorithms needs to be usable by everyone and does not need to use any shared code or if it is using that needs to be part of the mediator or it knees to be a external library that can be installed easily. At least that is my opinion, we could discuss more if you want.

[billsacks]

Thanks @uturuncoglu . I'd like to talk with you and @mvertens to work out the details of what this should look like. Regarding your point about the share code giving garbage when not initialized, I completely agree that there should be better checking there; I have just opened an issue for this: ESCOMP/CESM_share#83