SWMT shifts to lighter densities in MOM6-01 #40
Comments
|
We were wondering whether that had something to do with the initial stratification, so I looked at bottom T/S/rho and compared panant-01-zstar-v13 and panant-01-zstar-ACCESSyr2. All plots and more detailed text are here: https://github.com/schmidt-christina/mom6_comparison/blob/main/Shelf_properties_Plots.ipynb Initial temperatures don't differ much between panant-01-zstar-v13 and panant-01-zstar-ACCESSyr2 in the DSW formation regions, but it's up to 2°C warmer in the Amundsen/Bellingshausen Sea (not shown here). Salinity: The initial conditions in the Weddell Sea are fresher (up to 0.2 psu) in panant-01-zstar-v13 compared to panant-01-zstar-ACCESSyr2. But this reverses with panant-01-zstar-v13 being saltier everywhere on the shelf after 10 years. The Weddell Sea gets saltier in panant-01-zstar-v13 and fresher in panant-01-zstar-ACCESS-yr2. The Amundsen Sea gets saltier and the Ross Sea fresher in both runs, but the magnitude of the freshening is slightly stronger in panant-01-zstar-ACCESSyr2, so that panant-01-zstar-v13 is still saltier. Overall, it’s getting lighter everywhere on the Antarctic shelf in panant-01-zstar-ACCESS-yr2, especially in the Ross Sea.
Horizontal mean density profiles in the Weddell and Ross Sea have mostly the same shape in both runs, panant-01-zstar-ACCESS-yr2 is often lighter (with a constant offset over the whole water column compared to panant-01-zstar-v13). But it is not necessarily like that from the start, it often gets lighter over time as the SWMT shifts to lighter densities.
|
|
Amazing analysis, thanks @schmidt-christina! Rather unexpected, because we have the same atmospheric forcing, same vertical grid and same horizontal resolution / bathymetry as ACCESS-OM2-01. @AndyHoggANU suggested we could try playing around with the boundary layer vertical mixing schemes. I'll try to get some panan-01 test cases going today with different mixing schemes / parameters. |
|
Yeah, I agree this is a very nice analysis. I'm also very interested in seeing what we get out of the 1/20° case. |
|
I have 5 years of the 1/20th run now. So far it doesn't look too bad, there is certainly not the same shift that I can see already in the first 5 years in the 1/10th run with same IC + BC etc. But as I only have five years yet, I would still be a bit cautious. Also the SWMT is quite weak in the Weddell Sea.
Plot below is the panant-01-zstar-ACCESS-yr2 with the shift visible already in the first 5 years for comparison:
|
|
That’s good news! Did you decide at the meeting today to run longer? If so,
has anyone checked what the density binning on vmo and umo is like? Might
be good to double check the diagnostic list, so we don’t have to rerun 10
years (perhaps you already did this at the meeting today). Do we have x and
y heat flux diagnostics switched on too?
…On Mon, Apr 17, 2023 at 8:36 AM, Christina Schmidt ***@***.***> wrote:
I have 5 years of the 1/20th run now. So far it doesn't look too bad,
there is certainly not the same shift that I can see already in the first 5
years in the 1/10th run with same IC + BC etc. But as I only have five
years yet, I would still be a bit cautious. Also the SWMT is quite weak in
the Weddell Sea.
[image: image]
<https://user-images.githubusercontent.com/38216156/232346051-52918744-5362-497a-b61b-ed8c96901109.png>
Plot below is the panant-01-zstar-ACCESS-yr2 with the shift visible
already in the first 5 years for comparison:
[image: image]
<https://user-images.githubusercontent.com/38216156/232346063-d7b7dc9c-d4ad-463c-85ba-6cdeb1f76011.png>
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44UZCFUHRWX3DV3FSQ5TXBRYABANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
|
Yes, we decided to run another 5 years and it's running atm. We briefly looked at the density binning, but decided that I would look at it in more detail. So maybe I can look at it first thing tomorrow morning and pause the run in the meantime? What do you mean by "x and y heat flux diagnostics"? |
|
I wouldn't bother pausing, just keep running and see if you can check the
density binning as soon as you can.
Zonal and meridional heat fluxes, I'm not sure what they're called in MOM6.
WIlton, do you know, or could you check?
…On Mon, 17 Apr 2023 at 16:04, Christina Schmidt ***@***.***> wrote:
Yes, we decided to run another 5 years and it's running atm. We briefly
looked at the density binning, but decided that I would look at it in more
detail. So maybe I can look at it first thing tomorrow morning and pause
the run in the meantime?
What do you mean by "x and y heat flux diagnostics"?
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44U6QH6QSJMULF7WNRQ3XBTMP3ANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you commented.Message ID:
***@***.***>
|
|
I found this variable, is that the right one @adele157 @willaguiar ? |
|
That’s 3D convergence, which would work if we only want circumpolar
integrated heat fluxes. Better if we can have the actual x and y fluxes
though.
…On Tue, Apr 18, 2023 at 9:11 AM, Christina Schmidt ***@***.***> wrote:
I found this variable, is that the right one @adele157
<https://github.com/adele157> @willaguiar <https://github.com/willaguiar>
?
T_advection_xy" [Unused]
! modules: {ocean_model,ocean_model_z,ocean_model_rho2,ocean_model_d2,ocean_model_z_d2,ocean_model_rho2_d2}
! long_name: Horizontal convergence of residual mean advective fluxes of heat
! units: W m-2
! cell_methods: xh:mean yh:mean zl:sum area:mean
! variants: {T_advection_xy,T_advection_xy_xyave}
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44U5LH4REGSTZ5IKNOETXBXEZPANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
I believe the fields we want are T_adx, and T_ady |
from: https://github.com/NOAA-GFDL/MOM6-examples/blob/dev/gfdl/ice_ocean_SIS2/OM4_025/diag_table.MOM6 |
|
My tests in panan-01 of different surface boundary mixing schemes made zero difference (at least to these overturning diagnostics - I haven't checked SWMT). I started from the orange config and:
|
|
The same information, but now with a much better diagnostic - cross 1000m isobath transport (thanks to @schmidt-christina for working on the code):
We haven't made a contour for the 1/20th run yet, but would be good to compare that also! And maybe |
|
Turns out we don't have salinity restoring turned on in any of the panan configs. This is quite likely the problem! |
|
Restoring doesn't help. :( This is using the same restoring file as ACCESS-OM2-01 with the same piston velocity and I've checked it's actually doing something. |
|
Hmm. Can you describe how you calculated the piston velocity? |
|
MOM6 uses piston velocity as the input parameter: This matches what we use in ACCESS-OM2-01: |
|
Does MOM6 restores salinity under the ice as in MOM5 or has a different flag for it? Wondering if that could be it |
|
Yes, it also has an option to turn off restoring under ice. I opted to do
exactly what we do in ACCESS (restore under ice).
…On Thu, 27 Apr 2023 at 17:02, Wilton Aguiar ***@***.***> wrote:
Does MOM6 restores salinity under the ice as in MOM5 or has a different
flag for it? Wondering if that could be it
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44UZWDWJQ4MSRKKCVLUDXDIKZDANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
Poring through MOM6_parameter_doc.all, I noticed:
|
|
So, try a run with river mix? |
|
Yep can do.
…On Thu, 27 Apr 2023 at 21:20, Andy Hogg ***@***.***> wrote:
So, try a run with river mix?
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44U6APKYQDFNGI662SM3XDJI6LANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
Plan from this morning's meeting (feel free to edit this comment to add your name to any of these or to add more suggestions): 1. More diagnostic analysis of panan vs ACCESS-OM2-01 spinup:
Names of experiments to compare for the above analysis:
2. More detailed look at salinity restoring files. Salinity restoring using WOA data did not help DSW formation. Is this because WOA salinity is way too fresh in winter? Compare maps of WOA salinity restoring file ( 3. Check input and parameter files for any anomalies.
4. We noticed that 5. Salt in sea ice. I also noticed that we have |
|
Also the updated DSW export (across 1000m isobath) plot with restoring and rivermix cases:
I'm still running rivermix, but it's not looking hopeful. We can update this once we have 1/20th and ACCESS-OM2-01 data to compare also. |
|
I created the 1000 m isobath for the 1/20 model and computed the cross 1000m isobath transport (green line). The offshore transport is higher in the 1/20th simulation compared to the 1/10th simulation. In agreement with the SWMT analysis it is stable for the first 5 years and then starts to slowly decline in year 1996+1997.
|
|
I also used a lighter density threshold to calculate the cross 1000 m isobath transport for all 1/10th simulations to see if we still produce DSW but at lighter densities. This is the case for the runs with year ~200 ACCESS at northern boundary and init. cond, but not when we use year 2 ACCESS BC + IC.
|
|
Thanks @schmidt-christina -- very nice! Just to check: Is this 1/20° run starting from the same WOA restart as the |
|
Yes I think it’s definitely worth investigating the comparison of salinity
restoring data choices suggested above. If we come up with better data we
could do a quick test on the 1/10th first?
For initial conditions, WOA (which is pretty close to what the 1/20th is
currently starting from) is pretty terrible on the shelf. Maybe worth
comparing later years from ACCESS (I.e. v13 initial conditions) with WOA
and Schmidtko bottom data to see why the v13 case is better?
…On Thu, May 4, 2023 at 9:40 PM, Andy Hogg ***@***.***> wrote:
Thanks @schmidt-christina <https://github.com/schmidt-christina> -- very
nice! Just to check: Is this 1/20° run starting from the same WOA restart
as the ACCESSyr2 simulation?
If yes, my feeling is that we have some cause for optimism that the 1/20°
run will give us much better results than the 1/10° ... Is there an
appetite (and spare SU) to restart the 1/20° with salinity restoring fixed,
and any improvements we can make to the initial condition??
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44U2CRK7FCEYYMT4HTZDXEOIR5ANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
@AndyHoggANU: Yes, the IC + BC of the 1/20th run are identically to the ACCESSyr2 simulation I would also prefer tests in the 1/10th first, as it takes quite long to run the 1/20th (and is expensive), we would probably need at least 8 years to see changes. |
|
Great, Christina, do you or someone else want to look at possible restoring
data and put a 1/10th test on today, so we have something to look at on
Monday? I'm still travelling, so can't do anything until next week.
…On Fri, 5 May 2023 at 08:14, Christina Schmidt ***@***.***> wrote:
@AndyHoggANU <https://github.com/AndyHoggANU>: Yes, the IC + BC of the
1/20th run are identically to the ACCESSyr2 simulation
I would also prefer tests in the 1/10th first, as it takes quite long to
run the 1/20th (and is expensive), we would probably need at least 8 years
to see changes.
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ACA44UYBDFNSWHB2IBKFGX3XEQS47ANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
|
I can put the 1/10th to run today with restoring salinity + change in sea ice parameters (Bulk salinity, density, etc) . Do we have the IC+BC for year 2 and 200 of ACCESS ready for the panan01 or do we create those ? If we have, what are their dirs? |
|
I'd suggest just running just the case with yr 2 boundary forcing to start, i.e. a copy of this case: For the salinity restoring, have you compared winter maps of the 3 suggested salt data possibilities (ACCESS-OM2-01, Pauthenet, WOA) above to see which one will be best? Would be great to see those maps here (or a new issue perhaps - this one is getting long). |
|
If you want to see Salinity Anomalies in other seasons check here. |
|
I'm commenting on parameters here, as it seems a better spot for it #42 |
|
|
|
Ok, some plots comparing the SSS in WOA salinity restoring target with final Surface_salt from Fig 1 = Salinity anomalies of OM2 output - WOA reference, for June, first 2 years of
|
|
I don't really agree with your comment that output SSS in OM2 is fresher over the whole shelf in Jun than in WOA. It is certainly fresher very close to the coast, especially in the Weddell/Bellingshausen/Amundsen Seas, but that's not were we have strong SWMT. We normally have strong SWMT in the western Ross Sea and Adelie Coast, where the SSS in ACCESS-OM2 is a lot saltier , and in the western Weddell Sea (but not directly on the coast of the peninsula) and Prydz Bay, where the anomaly is not as strong, but it's still saltier in winter and spring. |
|
Thanks for plotting all these up @willaguiar! It's not as straightforward as I was thinking it might be. That really fresh band along the coast in ACCESS-OM2 might do weird things (make the shelf fresher) if we try restoring to that. But then again WOA is saltier, as Christina points out, directly over the DSW formation sites. It's not clear to me what the best test would be. I'm thinking maybe another restoring run using WOA data, but ramp up the restoring parameters. For the last test I used: We could try as above, but increasing these to the OM4 values of: To make this test a bit different to the last one, we could also use Pauthenet data south of 60S, patched to WOA north of that. It looks like Pauthenet is fairly different in the Amundsen / Bellingshausen / WAP, where we normally have large biases in ACCESS-OM2. Any other opinions? |
|
Also for those who missed yesterday's meeting, our plan for the next week is as follows:
Hopefully one of these tests might give us some better DSW formation in the 1/10th. But if not, we will start the 1/40th run next week anyway. |
|
I ran 10 years of panan-01 starting with initial conditions from "World ocean circulation experiment-argo global hydrographic climatology (WAGHC)" (Gouretski, 2018, Ocean Science). This run is probably even worse compared to starting from ACCESSyr2 IC by just looking at the cross-1000m transport but it does confirm that the initial conditions indeed have quite a strong effect on the DSW formation and export.
|
|
And results here for the test run with altered sea ice parameters (red line compared to equivalent control in orange): I changed a bunch of sea ice parameters, and the shortwave penetration depth, as outlined in this issue. I guess - while not helpful for our purposes right now in setting up the 1/40th - this does indicate there is a strong sensitivity to sea ice parameters. Maybe if we tried to change those parameters individually we'd be able to identify which are the crucial ones and modify them to get more DSW formation rather than less as we've done here. |
|
Added to Adeles plot here the case of restoring salinity with a stronger piston velocity ( 
I checked the salinity over the shelf (south of 1000m isobath), and both SSS and SOBottom get lower and lower within 10 years (it maybe explains why we have less and less DSW?) , while in a OM2 run restored towards WOA (IC from WOA) salinities over the shelf get roughly stable around 5yrs (Added there for comparrison) 
|
|
Turns out we do get some improvement in DSW formation in the panan-01 when we use these salinity restoring settings (green line compared with control orange line):
Note we also used Pauthenet et al. 2021 SSS instead of WOA, but I don't think that was the important part. The crucial parts seem to be We're running with these settings for the panan production runs now. |
|
nice. that's starting to make sense :)
P
…On Fri, May 26, 2023 at 11:28 AM Adele Morrison ***@***.***> wrote:
Turns out we do get some improvement in DSW formation in the panan-01 when
we use these salinity restoring settings (green line compared with control
orange line):
RESTORE_SALINITY = True
ADJUST_NET_SRESTORE_TO_ZERO = False
FLUXCONST_SALT = 0.1667
SRESTORE_AS_SFLUX = True
MAX_DELTA_SRESTORE = 5
[image: Screen Shot 2023-05-26 at 11 41 28 am]
<https://user-images.githubusercontent.com/8506963/241111121-2252cd0f-9bf2-4e11-842e-5ab88dfa86e3.png>
Note we also used Pauthenet et al. 2021 SSS instead of WOA, but I don't
think that was the important part. The crucial parts seem to be ADJUST_NET_SRESTORE_TO_ZERO
= False and MAX_DELTA_SRESTORE = 5. I think with all the Antarctic runoff
input at the surface next to the coast, we get a really extreme coastal
freshening. Ramping up MAX_DELTA_SRESTORE = 5 (same as GFDL OM4 setting)
acts to counteract the coastal freshening. Wilton's previous case had the
same high settings, but without ADJUST_NET_SRESTORE_TO_ZERO = False and
that did next to nothing, so I think with the global offset to zero the
restoring, we must have ended up putting the wrong sign restoring flux in
some places.
We're running with these settings for the panan production runs now.
—
Reply to this email directly, view it on GitHub
<#40 (comment)>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/ABSWJXFKFBEU4CD5W2GFOCLXIAPOFANCNFSM6AAAAAAW4S7JMY>
.
You are receiving this because you are subscribed to this thread.Message
ID: ***@***.***>
--
*I’m sending this message at a time that suits me. I don’t expect you to
reply outside of your own work hours.*
Paul Spence, Assoc. Prof.
ARC Future Fellow
Institute for Marine and Antarctic Studies <https://www.imas.utas.edu.au/>
University of Tasmania, Hobart, Australia
https://paulspence.github.io/
|
I calculated the SWMT in six different 1/10th MOM6 configurations:
The peak of the mean (1996-2000) SWMT in each of the 4 formation regions occurs at different densities with runs that start from year 200 of ACCESS-OM2-01 being denser (blue, orange, red line) and runs starting from year 2 of ACCESS-OM2-01 (i.e. close to WOA observations) being lighter. This is especially visible in the Weddell Sea.
The peak of the SWMT shifts to lighter densities each year in the runs that start from ACCESS-OM2-01 year 2 in all regions and especially in the Ross Sea.
In runs starting from ACCESS-OM2-01 year 200 the density of the peak SWMT isn't shifting to lighter densities as much and also slower (note that the 2nd Plot with panant-01-zstar-v13 includes 46 years more)
A timeries of the SWMT for a density at a mean (1991-2012) transport of 70% of the maximum transport (orange line) for panant-01-zstar-ACCESSyr2 shows that after 10 years the SWMT dropped below 0.5 Sv from around 2 Sv in the Weddell/Ross Seas.
I didn't see a difference between vertical coordinates, or whether it was global or panantarctic.
Plots for all of the runs are here: https://github.com/schmidt-christina/mom6_comparison/blob/main/SWMT_Plots.ipynb
The text was updated successfully, but these errors were encountered: