Adds internal ocean frazil ice porosity #6802
Conversation
These changes assume frazil ice retains salt until precipitating at the surface. Treatment under land ice remains the same. nonBFB
Adds namelist field config_frazil_ice_porosity Modifies the internal frazil salinity nonBFB
Depth of maximum frazil determined by config_frazil_maximum_depth Uses config_frazil_ice_porosity to compute internal frazil salinity nonBFB
|
|
Relevant discussion can be found at: |
|
|
|
Aach, sorry! I missed that that was linked above! |
|
@njeffery, how easy or hard would it be to run a model vs. model run of MPAS-Analysis so we could see the difference plots between your run (without BGC) and the control run? The differences are definitely there but they're subtle enough that those difference plots might really help. I'd be happy to help with running the analysis if need be. The main thing would be whether the analysis output (not just the plots) is still available for the control run. |
|
I have the first 100 years of the control run ocean history files only on Perlmutter: /global/cfs/projectdirs/e3sm/njeffery/E3SMv3/20231209.v3.LR.piControl-spinup.chrysalis/archive/ocn/hist . I could bring the ice history directory too. What else would you need? |
|
@njeffery, I think I would want everything on Chrysalis because that's where the control run is. Since you ran on Anvil, it's already there. I can give it a try tomorrow. The config files from your analysis run and Wuyin's should tell me where everything is (or was at least) and I can let you know if anything is missing. |
|
@njeffery, okay, I see the issue now. This run is no longer on Chrysalis. It is pretty difficult to compare runs on different machines so I think we would either need those 100 years of ocean and sea ice history, plus a restart file for ocean and sea ice, plus the namelist and streams files. Or we would need the MPAS-Analysis run that was already done (maybe easier). I will see if I can restore that with globus directly from tape at NERSC. Update:
|
|
@cbegeman, @darincomeau and @proteanplanet, I know you are both going to be busy with AGU but I wanted to check with you about this. I would not feel comfortable approving this change, which I think could be significant, without a B-case simulation with thermodynamics active in the ice-shelf cavities (PISMF - prognostic ice-shelf melt rates). I think this might mean doing a simulation branched of from @darincomeau's |
|
@njeffery, I now see that you've moved the first 100 years of data from your simulation to Perlmutter. I can do the analysis there then (on Monday). Could you give me permission to read the files and directories in the archive on your scratch drive? I'll do the same for my data once it's extracted. It's at: |
|
Thanks, @xylar. You should now have permission. |
|
@xylar and @njeffery: This is what has to happen to this PR from here in: It requires a B-case LR 100-year simulation to be compared with a control. Moreover, the change must be introduced initially to master as a BFB change. So the PR gets tested with the change to porosity to the value currently set in the code for this PR, but the code should be set with porosity at zero which I assume makes it BFB with master. Then, for the time being, we test, for example, SORRM and BGC configurations with the porosity set to non-zero in the namelist changes in run scripts, also the HR group, who are interested in testing this as well. Once we are all happy it's performing correctly, we then submit another PR making this default after broad consultation in group meetings. |
|
Documentation at the link provided by @njeffery meets requirements. |
I thought this initially as well but reading the code I think a porosity of zero would only be BFB under ice shelves, where the current salinity of ice is zero. In the open ocean, this is non-BFB in all conditions. |
|
I agree with @xylar in the open ocean the frazil reference salinity is used right now, so porosity of zero wouldn't be BFB in the open ocean. I also don't believe we can use the code as and just tuning the porosity to get BFB. |
|
Do you want me to add a config flag that adds an option which reverts to the original code? |
|
Sorry I was unclear, I'm requesting either a switch to make this BFB, which could be, for example, setting porosity to zero, or a separate flag. We are doing something similar for changes to top of atmosphere radiation for CMIP7, which will be highly likely to become the default, in a separate PR later. |
|
Sure. i can make it a stealth feature. |
|
One option, which may be along the lines that @proteanplanet was thinking, would be to have a config option that is the minimum frazil salinity, defaulting to 4, and then we could reproduce the current behavior of the model by setting frazil ice porosity to 0 (minimum would kick in everywhere). |
| <nml_option name="config_frazil_ice_porosity" type="real" default_value="0.85" units="1" | ||
| description="Assumed porosity of frazil" | ||
| possible_values="Any positive real number from 0 to 1." | ||
| /> |
There was a problem hiding this comment.
I'm wondering if we should call this something other than porosity. When I hear porosity, I assume that the volume of frazil ice formed is only porosity * frazil ice thickness tendency (* areaCell) because part of the frazil ice tendency (the volume advected upward) is pure seawater. But my understanding of the implementation is really that this is expressing the salinity of the frazil as a fraction of the surrounding seawater (the ice formed and latent heat extracted is the full frazil ice thickness tendency). So I wonder if a less confusing name would be something like config_frazil_ice_salinity_fraction.
There was a problem hiding this comment.
I also got a bit confused by the use of "porosity". I think the use of that term was partly responsible for the confusion we had in our webex discussion a little while back. I seemingly couldn't let go of the idea that not just salt but seawater was being captured and transported by frazil in this approach.
There was a problem hiding this comment.
@njeffery, do you have feedback on this?
|
A quick update on MPAS-Analysis: I have things all set on Perlmutter. I have submitted jobs but they're sitting in the queue. I'll post links here once they've run, hopefully tomorrow. |
Defines config_frazil_ice_porosity as: -Positive real number from 0 to 1 when used to define internal frazil salinity. -Negative value reverts to using the reference salinity for internal frazil. BFB. New frazil option included as stealth feature.
|
I've modified the code so that it's now BFB with default settings (verified in two short 10 day simulations). |
|
Regarding the MPAS-Analysis runs, I've been sick so not putting a huge amount of effort into the plotting but I'm also seeing long queue wait times and longer-than-expected run times for MPAS-Analysis on Perlmutter. Plus, I made a mistake with remapping that forced me to start over. I believe I should have the main and control runs plotted once my next pair of jobs runs. The main vs. ctrl plotting shouldn't take too long once that happens, since it's just a question of plotting without further computations. The Perlmutter experience does make me miss running analysis on Chrysalis. |
|
@xylar - I appreciate all your help! No rush and no worries. |
xylar
added
Stealth
Model vs. Model with latest MPAS-Analysis including "polar regions" analysisThis branchControl run@njeffery, could you add these links somewhere appropriate on the confluence page for the simulation? |
|
I'm mostly seeing ensemble-level variability but serious improvements in AMOC:
and quite a bit more ocean heat uptake in the North Atlantic:
Presumably related to more overall OHC: |
|
My brain picked up 'serious improvements in AMOC' and I immediately went to look at the diags :). A few other improvements I noticed are in the subpolar North Atlantic: more saline surface, which might lead to the more dense water mass formation in the Greenland Sea (from the TS diagrams), and also improved sea ice in the Lab Sea and Nordic Seas (as well as central Arctic ice thickness). This looks very promising! |
| <nml_option name="config_frazil_ice_density" type="real" default_value="1000.0" units="kg m^-3" | ||
| description="Assumed density of frazil. NOTE: test and make consistent with E3SM." | ||
| possible_values="Any positive real number." | ||
| /> | ||
| <nml_option name="config_frazil_ice_porosity" type="real" default_value="-1.0" units="1" | ||
| description="Internal porosity of frazil" | ||
| possible_values="Positive real number from 0 to 1. Negative value reverts to using the reference salinity for internal frazil." |
There was a problem hiding this comment.
| possible_values="Positive real number from 0 to 1. Negative value reverts to using the reference salinity for internal frazil." | |
| possible_values="Positive real number from 0 to 1. Indicates the fraction of local ocean salinity to use for frazil ice in the water column. If non-negative, applies to both land ice frazil and sea ice frazil. Negative value reverts to using the reference salinity for internal frazil and 0 for land ice. |
There was a problem hiding this comment.
@njeffery, I feel uncomfortable about this approach of using a negative value as a flag. This is not an approach we have taken elsewhere in MPAS-Ocean. Instead, we have, at least as far as I'm aware, always used a separate logical to turn the feature on or off.
However, if other reviewers are fine with this approach, I will defer to the majority.
There was a problem hiding this comment.
@xylar : that's a good point. I figured this would be a clean option because this is not really an independent feature. It just makes explicit the choice to use either a fixed reference salinity or a fraction of the local salinity in the internal frazil calculation.
There was a problem hiding this comment.
@njeffery, I take your point that it's a little different than a typical on/off feature. In such cases, we typically have a string namelist option that selects between the possible features and then we have one or more logicals in the code that determine the behavior (to avoid constantly performing string comparisons except at init within the code itself). Would that be an approach you could get behind here as well?
There was a problem hiding this comment.
Sure. I'll try to get to it this week.
|
Has there been any testing with land ice thermodynamics on (even in G-cases)? It would seem strange to me to merge untested changes that apply to land ice, even if this is a stealth feature right now. |
|
@cbegeman, I don't think it has been tested with active land-ice thermodynamics, no. Do you have a baseline you'd suggest for that? The BlueTip configurations might be too expensive to be attractive and they're all B-cases as far as I'm aware. |
| * config_frazil_ice_density | ||
| accumulatedFrazilIceSalinityNew(iCell) = accumulatedFrazilIceSalinityOld(iCell) + sumNewThicknessWeightedSaltContent | ||
| * config_frazil_ice_density | ||
|
|
There was a problem hiding this comment.
| ! accumulatedFrazilIceSalinityNew is overwritten below if usePorosity = 1 |
|
@njeffery Can you comment on the outcome of salt conservation as reflected in |
|
The lowSalinityFrazilIce diagnostic is nonzero when salinity conservation from frazil formation is violated. In all cases, this occurs when the cold ocean region is too fresh and doesn't have 4psu times the frazil mass to subtract from the ocean column. However when the sea ice in that region melts, the ocean assumes the melt water has 4 psu. The result is a small, additional positive salinity flux into the ocean.
|
Thanks! It's good to hear that the regions affected seem small. Am I correct in thinking that the salt non-conservation issue is also present in master, it just wasn't being explicitly tracked before? |
|
@cbegeman : That's right. This PR doesn't directly address the brackish water issue. |
|
The change in AMOC is definitely statistically significant and the standard deviation (noise) of the signal is reduced:
As a rough ballpark, it buys between 2 and 3 Sv based on results using 2-10 year low filtering. |
I think a G-case would be fine for a land ice thermodynamics test. @darincomeau What do you think about branching from year 21 of this case and running to year 30? Do we still have the restarts and the ability to do a MPAS-Analysis comparison? I checked that the land ice frazil accretion rates are non-negligible for this case https://acme-climate.atlassian.net/wiki/spaces/PSC/pages/4575821825/20240829.GMPAS-JRA1p5-DIB-PISMF.TL319_SOwISC12to30E3r3.mesh-review-test.chrysalis |
This scheme updates the internal frazil ice salinity redistribution. The default ocean frazil ice parametrization contributes to unphysical stratification of the upper ocean during formation because it uses the ocean-sea ice coupling salinity for internal ocean redistribution.
Changes and simulations are documented here: https://acme-climate.atlassian.net/wiki/spaces/HESF/pages/4820598785/Ocean+Frazil+Ice+Porosity
The code contains comments !NJ to help explain the changes. I can clean that up once the code has been review.
[NML]
[Stealth]