Low frequency variability in the Southern Ocean

[islasimpson]

Here is a place to discuss low frequency variability in the Southern Ocean in recent runs. @iangrooms @gustavo-marques @duvivier @adamrher

e.g., this low frequency variability in ACC transport that has been seen by @gustavo-marques and @iangrooms

[islasimpson]

Here is a look at TS in various piControls and 271 and 276. First, here's a simple analysis of TS standard deviation. I just omitted the first 50 years of each of them and plotted the 40-year running mean standard deviation. Lots of TS variance in the Southern Ocean in the Pacific sector in most of the runs, and perhaps more so in 271 and 276 as @adamrher was discussing in the context of trends in the project meeting. Of course we don't really know what we're looking at here e.g., is it differences in spin-up or actual low frequency variability.

Here's focussing on the blue box region in the Southern Ocean and showing annual mean timeseries and 20-year low pass filtered timeseries.

TS in this box looks quite highly correlated with the drake transport in the above figure. I think with this low frequency variability and with these short runs, it's hard to say whether we're looking at something really different here. I definitely think it's hard to say that 276 and 271 are different just from eyeballing.

Here's now TS anomalies in that box for the historicals and the equivalent part of the PI control. One thing to note here is that the 271 TS in the historical starts to turn around. Perhaps the same is true for the ACC transport. Clearly the variability that's present in the piControl, whether that be equilibration or internal variability in TS in this region is dominating in what happens over the historical record too.

Here's an attempt at providing some insight into whether we think such variability is undesirable or not. It's showing TS in the box from 1950 to 2020 over which time BEST has coverage in this region and comparing both the historical and the piControls with BEST . Of course we only have one observational sample and we're talking about low frequency variability so it's difficult. But I think you get the sense from this figure that the low frequency variability that we see in TS in this region in the simulations is bigger than what's present in the observations. The observations exhiit a relatively constant long-term trend that's smaller than the low frequency fluctuations we see in the simulations.

We can take a look at this when our historicals have runs to the end and it seems like it would be worth running the piControls out longer to see whether 271 starts to reverse again, like it's companion historical does.

[adamrher]

Thanks Isla for this analysis. Doesn't it seem like the large low frequency variability in 158 is stabilizing / not present after year 175? That wold suggest this is spin-up related. We're keeping the 271 & 276 PI controls going so we'll see how that looks as we enter the 3rd century.

[islasimpson]

Yeah, it does. Not sure how much money I'd bet on it staying that way though if the run were longer.

[islasimpson]

But Gokhan was arguing for a long-timescale equilibration yesterday based on what happened in the low res simulations of MESACLIP as in this figure from Chang et al...

[duvivier]

Just to keep the conversation in this thread. These figures are from this historical runs, but we see the SH sea ice oscillate in the PI control, but it's in the historical that it really nosedives in the early 1900s - way too early compared to obs - but then it recovers again. It does seem the main signal is not in the Ross Sea but in the Weddell Sea. This does appear to possibly be related to the Southern Ocean oscillation behavior Gokhan mentioned since it recovers.

However, it's worth noting that 276 historical Arctic sea ice also gets funky around the same time and that this would not be related to ocean ventilation in the Southern Ocean. In the Arctic the signal is not in concentration, but seen in thickness and volume where the thickness declines and volume drops. We have no idea where this is coming from and it also does not "recover" like the SH extent did.

[iangrooms]

We just had a meeting to discuss this. Here are a few points I remember.

• We think there are two things going on with the ACC: a declining trend and multidecadal variability. They are related but not the same.
• The declining trend may be associated with reduced wind stress over the Southern Ocean. We can try to increase ACC transport to offset this by reducing the GM coefficient. As a point of comparison, in run 259 the GM coefficient in 259 is reduced by about 1/3 compared to our more recent (and also our older) runs. The ACC transport in run 259 increased over about 50 years, which is in line with previous experience: reducing GM increases ACC transport. Unfortunately there were many detrimental impacts of reducing GM, including a general reduction of SST and a significant increase in sea ice, but we can try to reduce GM a bit more gently to try to bring overall ACC strength back up without damaging the other things. (This is not related to the oscillations/variability.)
• @gustavo-marques looked at runs 091 and 092 which only differed in the vertical coordinate. 091, with z-star, did not have ACC variability and 092, with hybrid coordinate, did (though at a lower amplitude than the variability we're seeing now). We have also already seen (see head of this discussion) that ACC variability is correlated with temperature anomalies at 2ish km deep south of 60S. The hybrid coordinate has less spurious vertical mixing than the z-star coordinate; this reduced mixing may allow the buildup of temperature anomalies at depth that reduce the density gradient across the ACC and reduce the transport. We might be able to offset this by increasing the resolved/physical mixing.

[iangrooms]

Here are plots of the Southern Ocean Deep Mixing Volume (DMV) and Drake Passage transport in 276 showing that ACC increases are associated with deep convection. Monthly DMV is the volume of water between the surface and the bottom of the mixed layer in regions where the mixed layer is at least 2000m deep; computed using monthly mlotst_max values. The first two pulses are in the Ross sector; the third pulse comes from the Ross and Weddell sectors.

[islasimpson]

Here are some plots looking at SH zonal wind stress and zonal winds in 156 compared to 271 (I don't have data for 128 which was the other run suggested).

Zonal mean surface wind stress. The position of the maximum zonal wind stress very similar between 156 and 271, except for in DJF where 271 is a bit equatorward. I'm not sure how much faith I would put in this comparison with the obs. I prefer comparing the near surface winds since it's not relying on the surface flux schemes to produce it and it should be somewhat constrained by geostrophic balance and satellite temperature measurements etc.

Here's the 950 hPa zonal wind. Notice that you draw a different conclusion as to which is better in terms of magnitude. The wind speed in 156 is actually closer to ERA5 than 271 (the opposite of the conclusion drawn with taux). In 271, the near surface zonal wind is slightly equatorward and slightly weaker than what we had in 156. I guess from the ocean perspective it's TAUX differences in the first plot that are the thing we care about, but I've added this 950 hPa U plot for the comparison to obs.

[islasimpson]

One other thing to note is that external forcings (GHG's and ozone) tend to shift the westerlies poleward an strengthen them, so even the U950 comparison isn't really that good of a comparison. I could look at historicals for a better comparison, but I don't think we have one for 156.

[islasimpson]

Here are the same plots but for 156 compared to 178. I think they look fairly similar to the differences between 156 and 271 but they're not exactly the same. In DJF, there's not a shift in the wind stress maximum latitude like there is above. There's also more of a strengthening of the wind stress maximum in each season here. In the above, the increase in the red one in DJF and MAM is less than it is here.

[JulioTBacmeister]

Going through cesm_dev issues I saw this in issue #90 . I have forgotten what we actually wound up doing with Beljaars and Antarctic SGH30, but it might be worth looking in to.

================== from issue #90
Same configuration as 164 (#88):

user_nl_cam

bnd_topo = '/glade/campaign/cgd/amp/pel/topo/cesm3/ne30pg3_gmted2010_modis_bedmachine_nc3000_Laplace0050_noleak_20250325.nc'

clubb_c8 = 4.5->4.6
SourceMods for beljaars
/glade/work/bramberg/cam6_4_063/cases/c64_63_ne30pg3_FMTHIST_eff025_hdepth032_allmods_roughtopo_beljaar/SourceMods/src.cam/beljaars_drag.F90

[danabasoglu]

[danabasoglu]

This is from my 2012 paper, showing the Drake Passage (ACC) transport from several simulations along with corresponding zonal-mean zonal wind stress. In these simulations, GMs were comparable, I believe. So, there is quite a bit of sensitivity to the zonal wind stress.

[JulioTBacmeister]

I think Beljaars has been multiplied a factor of 4x in 178 compared to 156. This is applied globally.

SGH30 in the experiments is the same, except for an enhancement applied around Greenland in 156.

Thus, Beljaars drag is generally 4x stronger in 178 than in in 156. This could have an impact on stress around the Southern Ocean.

[duvivier]

All - several of the sea ice folks have a bi-weekly Antarctic sea ice discussion. Today we discussed this problem. Will Hobbs mentioned that the ACCESS model, which uses MOM6, does NOT have the oscillation we're seeing and that GFDL also had. @iangrooms, @gustavo-marques have either of you or Gokhan spoken with the ACCESS team and looked into what their parameters are? Maybe this is barking up the wrong tree, but something we discussed for a while.

[danabasoglu]

We can double check, but I think they are using the z* formulation in the vertical. I will reach out to them to confirm.

[danabasoglu]

We got some information from the ACCESS side. They have order 50-year G-case simulations. They have their first coupled simulation which is only 50 years with "lots of issues to iron out". So, they cannot say if they have any ACC oscillations or not. These are all with MOM6 at 0.25-degree horizontal resolution. It is also confirmed that they are using z*.

At yesterday's section meeting, we decided to try a run with z* in MOM6 as a sensitivity simulation.

[duvivier]

@danabasoglu Thanks for the follow up. I wasn't sure if it would be useful or not, but wanted to at least let folks know about the discussion. We also discussed that we should maybe add a Weddell Sea regional sea ice area metric to the sea ice metrics (like we do for the Lab Sea) to verify if we're seeing actual polynyas every time this happens. I would guess we would see a signature if we look regionally, but we don't with the hemispheric timeseries.

[iangrooms]

Do we have any diagnostics for precipitation south of 60S? I am wondering if our Southern Ocean problem is analogous to our Lab Sea problem. In the Lab Sea cold fresh water overlies warm salty water, with the stable density stratification driven by salinity. In the northern hemisphere we have too much precipitation leading to too much fresh water in the surface ocean. That enhances the stratification in the Lab Sea, which can prevent convection from bringing the warm water up to melt sea ice. We have altered the ocean parameterizations in such a way that this cold fresh water stays more in the boundary current, with less of it entering the central Lab Sea where it might otherwise shut down convection.

In the Ross and Weddell seas cold fresh water similarly overlies warm salty water. An excess of precipitation over the southern ocean and Antarctica (e.g. south of 60S) could lead to too much cold/fresh water near the surface, enhancing the stratification, and preventing convection from releasing the deep heat anomalies. The buildup of these deep heat anomalies seems to be associated with the reduction of ACC strength (by reducing the meridional density gradient across the ACC). The buildup of heat weakens the stratification until eventually deep convection can penetrate and release it, allowing the ACC to recover.

If there is an excess of precipitation at high southern latitudes, and if it could be reduced by atmospheric model tuning, it might keep the surface stratification weak enough that regular convection could prevent the slow buildup of deep heat anomalies.

[islasimpson]

Here are some plots of 271 during the historical period (1979-2014) relative to GPCP. North of 60S we have a wet bias. Around 60S we have a dry bias and then south of that we have a wet bias. So it would depend a bit on exactly what latitude is important.

[adamrher]

@iangrooms you may want to keep an eye on this new run 297 (#277), which is a clone of 271 based on a suggestion from @JulioTBacmeister. We have been running with enhanced Beljaars roughness over Antarctica to reduce precipitation biases over the ice sheet. The reasoning is that the ice sheet margins are overly smoothed at 1 deg and cannot resolve orographic precipitation, permitting moisture to penetrate further into the interior, leading to +ve precip biases over the ice sheet. The enhanced Beljaars roughening facilitates convergence and upward motion at the ice margin, leading to precipitation occurring further upstream, with some of that now occurring over the ocean. Here's a plot of the impact of Beljaars roughening on Antarctica precipitation. The ocean is masked out, but you can see that columns 2 and 3, which have the enhanced roughening, are associated with negative -ve anomalies, suggesting there would now be +ve anomalies over the adjacent ocean points.