Missing variables in spinup outputs for seaice: sisnconc (monthly) and sisnthick (daily, monthly)

[rbeucher]

Missing variables in spinup outputs for seaice: sisnconc (monthly) and sisnthick (daily)
Maybe I am missing something @anton-seaice. CMIP7 requires sisnconc monthly and sisnthick (monthly and daily).

I checked the files in /g/data/p73/archive/CMIP7/ACCESS-ESM1-6/spinup/Dec25-PI-control/output283/ice

[anton-seaice]

That's correct - these are not prognostic on the sea ice grid in esm1.6

They are prognostic in the atmosphere model, I am not sure if we save them. (But they wouldn't meet the data request as they are on the atmosphere grid)

[rbeucher]

OK Thanks for clarifying @anton-seaice . I believe they are listed as core variables in the data request. Do we expect this to be a pb for submission? Maybe a question for Siobhan?

[anton-seaice]

My thinking was that we wouldn't submit them.

@ofa001 - did you have thoughts on this ?

The variables are snow (on sea ice) thickness and concentration

[ofa001]

Hi @anton-seaice I thought we had the snow thickness saved, that is a fairly standard variable in CICE, not sure if it was switched on in the daily output. The snow concentration was the tricky one we had to pull a formulae from the radiation code or just submit the ice concentration last time so assume its the same. We could drop it from the stats and let people who need it use their preferred formula from the other variables (normally needs hs though)

Looking back at some old runs pre the name changes to CMIP7 sisnthick variable names we were outputting daily hs (snow thickness) in the esm1.6 stats so when did we stop, it was even there in Sept/Oct when you switched over to the new names, in your test set ups. The 23October monthly one even had a formulae in it for snow concentration whilst the month I checked for the Sept run didn't look like it did.

The November stats its missing, how did I miss that. What did we change in the code, that caused it to fall over, I know we had another discussion about the formulae on snow concentration, but snow thickness should never have been dropped from the list.

I think we definitely need to provide snow thickness output its a basic climate variable, and a basic measure that's comparable to field measurements.

I will look at whats in the code and namelist next.

[ofa001]

Hi @anton-seaice looking at the namelist sisnconc and sisnthick have both been removed form it but I don't remember any discussion on removing the snow thickness perhaps my mind has yet to get in gear post Christmas. The snow concentration is controversial so that's Ok but the thickness is widely used. Do you remember what the problem was.

[anton-seaice]

sisnthick is asking for actual snow thickness averaged only over the snow covered part of the sea-ice whilst hs is an equivalent thickness (effectively snow volume divided by grid cell area)

So calculating sisnthick as a diagnostic could be hs/sisnconc but as sisnconc is not prognostic in CICE (in ESM1.6) it seemed incorrect to (re)calculate a snow concentration in CICE (e.g. there's no process in the model for it to be a diagnostic for).

Snow mass (sisnmass) is saved, which is closest to hs in this case (as snow density is fixed). Snow volumne (hs / could be sisnvol) is not requested.

[ofa001]

Question @anton-seaice what variable have other models been putting in that use CICE has there been discussions on this on the CICE pages. Through the CICE calculations we use hs and snow volume which we define from hsn and aicen in the different ice categories. I think last November you were going to send a query to SIMIP folk did you eve get a reply, I guess the reason snow thickness is relevan tis its down as one of the "essential climate variables" to be measure in the field, with more snow buoys and I guess more remote sensing techniques being used, I guess they can use the snow mass "sisinmass" and convert.

[anton-seaice]

CICE6 reports sisnconc using snowfrac - I think snowfrac is only impacted by meltponds and radiation - neither of which are in CICE in esm1.6

We could report sisnconc as 0 or 1 based on this from Notz et al 2016:

D1.5 Snow area fraction (sisnconc) Area fraction of the sea-ice surface that is covered by snow. In many models that do not explicitly resolve an areal fraction of snow, this variable will always be either 0 or 1.

sisnthick could then be vsno/aice where vsno>0 - e.g. assume a uniform snow thickness over all sea-ice. (this would equal $\frac{sisnmass}{317*(siconc/100)}$ ).

[ofa001]

Hi @anton-seaice thats what we did last time in CMIP6 and would be the best comprimise for ESM1.6. For CM3 we willl have meltponds so can calculate a snow fraction based on that in the code, we didn't save it in the history it during the CM2 code but its easy to add. If we consider the radiation effects as well those terms are done in the UM, which used the compound formulae that @blimlim mentioned in a github page on CM3 which had the combination from JULES code of the snow and meltpond effect overlapping on (albedo) we should check how that plays in to this on snow cover when we get to thinking how to a snow fraction it for CM3.

Lets stick with the stick simpler Notz approach for ESM1.6 and just treat it as either 0 or 1 of the ice concentration.

[anton-seaice]

It's still confusing though, if we report sisnconc as 0 or 1 then it's not correct for radiation I guess? Radiation code in the UM presumably considers sisnconc when calculating sea-ice thermodynamic fluxes?

[blimlim]

Radiation code in the UM presumably considers sisnconc when calculating sea-ice thermodynamic fluxes?

Yeah, the ice albedo gets set using the calculation here: https://github.com/ACCESS-NRI/UM7/blob/6602dadd15c190ee37c6644190f52d428bc66917/umbase_hg3/src/atmosphere/short_wave_radiation/ftsa.F90#L201-L202

which we tried to replicate in CICE http://github.com/ACCESS-NRI/cice5/blob/access-esm1.6/drivers/access/CICE_RunMod.F90#L529-L535 but I think we weren't too sure about including it

[ofa001]

Hi @bimlim I did wonder too about that formulae and its cross reference to the UM code, it sounded like it was also using formulae (maskd) that had been used in the land part of the Jules/Moses code as well I haven't checked that.

But it is a strange form from a CICE point of view the more normal one was the one form the NCAR radiation code where there was the snowpatch and hs snow thickness used but this we now introduce a double dependence as we are now suggesting snow thickness should be dependent on snow concentration :(

@anton-seaice
is right though we do have an inconsistency possibly one of many between the UM and CICE code, its certainly an issue in how its calulating the relative snow and ice albedo which is relatted to the snow fraction other thermodynamic fluxes will be more dependent on the ice area.

[anton-seaice]

CICE-Consortium/CICE#1066 (comment) is relevant:

Ah, so it might need to be divided by snowfrac as well. That is different. This is a rabbit hole that I was trying to avoid. The issue is that snowfrac is only the radiative snow fraction and does not apply to snow thickness in the thermodynamics. So, I would prefer to output snow volume / sea ice area (which assumes a snow fraction of 1). This is similar to the pond variables. However, pond fraction and thickness are both only used in the radiation. So, I was comfortable doing the correct division there.

[ofa001]

Looks like they are in the midst of the same conversation. :)