E3SM-Project
Change snow shape default to hexagonal_plate
This PR changes the default snow grain shape for all configurations to hexagonal_plate, which is used by the SNICAR-ADv3 model to scatter light. Thus, this option has an appreciable impact on albedo in snow-covered regions. The current default is sphere, thus this change increases albedo in snow-covered regions (aspect ratio of snow grains increases from 1 to 2.5).
This change is motivated by a significant albedo bias over polar ice sheets and a related summertime warm surface air temperature bias.
This change is supported by hexagonal_plate being the default in CLM5.
[CC]
The hexagonal_plate option was added in https://github.com/E3SM-Project/E3SM/pull/5102. It was assessed in this manuscript by @daleihao and others: https://gmd.copernicus.org/articles/16/75/2023/gmd-16-75-2023.pdf
Testing: v3 LR
This option was run in a 100-year B-case comparable with v3.LR.piControl (v3.0.2) https://acme-climate.atlassian.net/wiki/spaces/CM/pages/5133533185/V3.LR.piControl.hexagonal.
Top row: The albedo differences (v3.LR.piControl.hexagonal minus v3.LR.piControl) are on the order of +0.1, consistent with Dalei Hao et al. (2023).
Bottom row: The Tref differences are on the order of minus 3-5 degC over polar ice sheets.
First column: snow_shape = hexagonal_plate
Second column: snow_shape = sphere
Top row: DJF albedo difference from CERES-EBAFv41
Bottom row: JJA albedo difference from CERES-EBAFv41
Top row: DJF Tref difference from MERRA2
Bottom row: JJA Tref difference from MERRA2
Top row: DJF albedo difference from MERRA2
Bottom row: DJF Tref difference from MERRA2
OHC trends are more negative (appropriate for piControl conditions), with most cooling in the Southern Ocean where the piControl run has a SST warm bias.
The Arctic snow volume increases over the course of the simulation:
This is believed to be due both to enhanced albedo of the Greenland Ice Sheet and enhanced albedo from late-season snow cover in the Canadian archipelago (Fractional difference in snow cover in the month of June from years 85-95):
The increase in SH sea ice volume is more muted:
For more figures, see the simulation page where a powerpoint presentation given to the Coupled group is also linked.
@czender @bishtgautam and @daleihao Do you have any feedback on this PR? In particular, I was wondering if there was a longer ELM test that is needed to evaluate spin-up of the snow pack.
@proteanplanet Let me know if you would like me to provide further details here. In particular, would you like me to include the MPAS-Seaice material from last week's presentation or would you like to write up those findings yourself?
Testing: v3 HR
5-years of the v3 HR configuration have been run with this option. The results are consistent with v3 LR. This comment will be updated as more results are available.
DJF Albedo vs. CERES
JJA Albedo vs. CERES
DJF Tref vs. MERRA2
JJA Tref vs. MERRA2
@cbegeman Thanks for leading this effort. The 100yr LR and 5 yr HR simulation seem sufficient to me. I do not know of any other tests that would need to be performed prior to acceptance/merging.
@czender @bishtgautam and @daleihao Do you have any feedback on this PR? In particular, I was wondering if there was a longer ELM test that is needed to evaluate spin-up of the snow pack.
Happy to see this effort. The existing simulaitons look good to me.
@cbegeman Yes, please add information about the increase snow, especially in the Archipelago, and the change in both northern and southern hemisphere sea ice volume. We need this documented as part of the PR. Once that's in place, I think we are ready to take this live.
@proteanplanet The comment above has been updated to include sea ice volume time series and a June climatology of snow cover (the month for which the fractional difference is greatest in the Canadian archipelago). Let me know if there are any other figures you would like highlighted here.
This change (left column below) also had the desired impact in the SORRMv3 configuration of reducing Antarctic ice sheet surface albedo and temperature biases:
Please @njeffery, @xuezhengllnl, @bpbond add your review if you believe this is ready to become the E3SM default.
@proteanplanet and @cbegeman : Review is coming shortly. I've completed a 20-year fully-coupled bgc run with the hex option. Just waiting on some analysis.
@cbegeman : I've tested with a 20-year Pre-Industrial fully-coupled BGC simulation. Observed similar increases in NH/SH sea ice thickness/extent and cooling overall in polar regions. This change will likely make the Labrador and Weddell sea ice biases worse in the historical. Though there are definitely several areas (Barents, Sea of Okhotsk, Indian Ocean Sector) that will likely be improved. I don't have any immediate concerns with the marine/ice BGC, though I'll continue the run. The changes are significant though and any use in a campaign will require a long spinup. Is this a concern for anyone?
@njeffery Thanks for that analysis. I'm not sure what you mean by "will require a long spinup." I would think the spin-up time would be similar whether hexagonal or spherical snow grain shapes were used. Do you mean having to redo spin-up?
@cbegeman : Agreed, for HES and SORRM, we both need a long spin-up with or without the hexagonal option. I was asking the question broadly in case HR or water cycle wanted to use the option.
@njeffery Got it. And HR is planning to use this feature and has started using it in our tests.
The MVK test was the only one of the non-bfb tests that showed a climate diff: https://web.lcrc.anl.gov/public/e3sm/e3smtest/evv/MVK_PS.ne4pg2_oQU480.F2010.chrysalis_intel.C.JNextNbfb20250702_010317/validation/JNextNbfb20250702_010317.html
The MVK test was the only one of the non-bfb tests that showed a climate diff: https://web.lcrc.anl.gov/public/e3sm/e3smtest/evv/MVK_PS.ne4pg2_oQU480.F2010.chrysalis_intel.C.JNextNbfb20250702_010317/validation/JNextNbfb20250702_010317.html
I took a look at the failed fields and the biggest differences are what we would expect from this change; snow depth and reference height temperature.
@rljacob, am I right in thinking that your bigger concern is understanding why the other tests meant to detect climate-changing results (PGN and TSC, https://my.cdash.org/viewTest.php?buildid=2979932) didn't show climate-changing results? That's what it sounded like on Slack.
The gist of our slack discussion is that it is not worrisome that MVK fails while the other tests pass, as MVK is a more general test and runs the model for a longer period of time. Thanks for contextualizing @singhbalwinder, @bishtgautam and @whannah1!
@xylar yes I was curious about that but I think we came to an understanding on slack: PGN and TSC don't run long enough and are to atmosphere-focused to show an effect from this change.