smb_input=1 can cause CISM to only use SMB in the input file #130
Description
Currently, cism_wrapper and CISM support two values for the smb_input configuration:
0: SMB input in units of m/yr ice (same as acab) 1: SMB input in units of mm/yr water equivalent Default: 1 (but current CISM default is 0)
However, if the value is set to 1, we have discovered that this causes CISM to read SMB from the input file, or simply ignore the coupler input data for SMB. From an email with @whlipscomb
I think the problem may be with the config setting smb_input. This setting tells CISM (in the Glissade driver) whether the input SMB has units of mm/yr water equivalent (the units of model%climate%smb) or m/s ice (the units of model%climate%acab). If smb_input = 1, Glissade will compute acab by copying from smb and changing the units. If smb_input = 0, Glissade will ignore the smb field and use the existing value of acab.
In standalone CISM, this means we should have smb_input = 0 when reading acab from the input file, and smb_input = 1 when reading smb from the input file.
But in coupled CESM, it's not so simple. That's because we have Glad code in between. While the incoming qsmb from the coupled has units of mm/yr w.e., Glad converts to acab units when it's filling the arrays params%instances%acab and model%climate%acab. So when we get to Glissade, we should not compute model%climate%acab from model%climate%smb, but rather should go with the existing model%climate%acab, i.e. we want smb_input = 0.
[If you want to see the details, take a look at glad_main.F90, starting around l. 800. This is where Glad converts units. The qsmb field comes from the coupler with units of mm/yr w.e. (or equivalently, kg/m2/s) and is copied into params%instances(i)%acab. Then at l. 822, the units are converted to m/yr ice just before the call to glad_i_tstep_gcm. Subroutine glad_i_tstep_gcm is in glad_timestep.F90. At l. 200 there is a call to glide_set_acab. This is where the acab value in the params%instances derived type is copied into model%climate%acab. Subroutine glide_set_acab is in glide_io.F90. It converts the units from m/yr ice to m/s ice, which are the units CISM now uses.]
So I think the near-term fix is to set smb_input = 0 in cism_wrapper. You can check whether this was true for previous coupled runs.
For the next tag (cism_wrap_2_2_013), I am going to remove the option to use "1" in CESM because it seems to not work as expected at all. This issue describes the complicated coding and some of the unit change issues that should probably be refactored, so I am leaving it as future work.
We should also add a test in CISM_Wrapper to ensure that the coupler fluxes are being correctly read by CISM. I'm envisioning a test that looks for the input data files in a Tcompset and compares to history output from CISM to make sure that the right values are being used. Code needed for this would be a new test type (Probably a derivative of SMS tests) and history options from CISM outputing the given SMB.
@gunterl Did a great job discovering this issue in cism_wrap_2_2_012.