Why does the Amundsen get so cold?

[adele-morrison]

As shown here, we see a strong cold bias growing in the Amundsen sector on the shelf. This extends from mid-depth down to the bottom. What's causing this?

The temperature transect and the reduction in age in the basal run suggest we may be getting convection there that didn't occur in the control. The control may have been strongly capped by surface freshwater, which increased the stability. Now we put the runoff at depth the water column is more unstable and prone to convection.

Let's plot the change in mixed layer depth in West Antarctica and see if this is consistent with this hypothesis.

[adele-morrison]

Yes, mixed layers deepen a lot in the Amundsen region. The left plots show September MLD averaged over years 5-10. The right plots show the maximum daily MLD over years 5-10. All values are scaled to show MLD or change in MLD as a percentage of the local ocean depth.

[pedrocol]

The cold bias is due to the gade-line formulation, in which the latent heat of fusion is accounted. The control run input the runoff with SST

The parameterization seems to work fine in the initial years, specially for the Basal run. We can see a well formed plume in the age tracer for the year mean of the 2nd year.

But in year 4 (annual mean) we can already see that the convection affects the plume

[adele-morrison]

Pedro, are you saying that the Amundsen convection is caused by the additional cooling of the basal heat flux? It's not obvious to me whether it's caused by the heat flux change or the change in stratification from the freshwater input at depth making the water column less stable. I guess if we did the extra coding for the surface water mass transformation diagnostic we could quantify this.

On Fri, 11 Aug 2023 at 12:08, Pedro @.***> wrote:

The cold bias is due to the gade-line formulation, in which the latent heat of fusion is accounted. The control run input the runoff with SST

The parameterization seems to work fine in the initial years, specially for the Basal run. We can see a well formed plume in the age tracer for the year mean of the 2nd year.

[image: age_y2] https://user-images.githubusercontent.com/23285319/259907559-45cf25e7-cca7-4f20-acd3-80e256f498a1.png

But in year 4 (annual mean) we can already see that the convection affects the plume

[image: age_y4] https://user-images.githubusercontent.com/23285319/259908027-165aa607-a916-47f9-95ab-188fd8e0ee9c.png

— Reply to this email directly, view it on GitHub https://github.com/pedrocol/basal_mom5-collaborative-project/issues/69#issuecomment-1674134892, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACA44UZ2JQYNUKISPDDV73TXUWHX3ANCNFSM6AAAAAA3KYKTG4 . You are receiving this because you authored the thread.Message ID: @.*** com>

[pedrocol]

I think it is caused by the additional cooling, but not directly. I think that the cooling at depth, decrease the temperature on the shelf, until reaching the freezing point temperature. At that point, sea-ice is formed due to processes happening at depth, with a consequent brine rejection, producing denser water coming from the surface, which ends up in a water column less stable. Below a plot of temp, salt and age anomalies in July of the second year. We can see how the brine rejection blocks the cold and new water in such an early stage of the run.

For some reason, this additional cooling doesn't leaves the shelf. We could run an extra test with no cooling to check this.

[adele-morrison]

Things to investigate:

• 3D spatial distribution of supercooled waters
• 3D frazil production distribution (may need extra output)
• sea ice production rates in problem regions
• Check for numerical noise in any of the above.
• Compute freezing point in the same way that the model does (pre-TEOS-10) for the relevant salinities and pressures where the supercooled waters are occurring.

[aekiss]

At all three resolutions tracers (salt, temperature and age) are advected horizontally and ver- tically by the multidimensional piecewise parabolic method (Colella and Woodward, 1984), with a monotonicity-preserving flux limiter following Suresh and Huynh (1997) (horizontal-advection-scheme=vertical- advection-scheme=’mdppm’ and ppm_hlimiter=ppm_vlimiter=3 for temp, salt and age in ocean/field_table)

[aekiss]

Frazil production is not confined to the top model layer (frazil_only_in_surface=false) and uses the pre-TEOS-10 freezing temperature (freezing_temp_preteos10)=true from Jackett et al 2006 http://dx.doi.org/10.1175/JTECH1946.1 which depends on salinity and pressure

[hrsdawson]

Things to investigate:

* 3D spatial distribution of supercooled waters

* 3D frazil production distribution (may need extra output)

Re spatial distribution of cold waters. These occur along coastlines from the Amundsen Sea to the west Antarctic Peninsula and also along portions of the East Antarctic coastline. Here I'll just show figures from the Amundsen and west Antarctic Peninsula regions.

In the Amundsen Sea, the cold waters predominantly occur from January to April (I've only look at year 10).

When plotted by longitude and depth we can see they are (in this region) mostly confined to the 50-350m depth range:

And plotted along a single latitude - the green line in the above spatial map:

I don't think we have the 3D frazil diagnostics (do we?), but I've looked at the frazil_3d_int_z diagnostic which is the "Vertical sum of ocn frazil heat flux over time step" in $W/m^2$. The 2D fields look similar, but in the two simulations with $T_{basal}$ on the Gade line, we get discrete coastal locations where the frazil heat flux is significantly higher than in the control. E.g. in May 1909 max values in control are 226 $W/m^2$ and in the Basal_Gade and ICB_Gade simulations the maximum values are 5,381 and 7,021 $W/m^2$ respectively. These locations, where frazil_3d_int_z is at least an order of magnitude higher than the control, are indicated by red dots. These don't always coincidence with locations where we get the super cold waters though, so I'm not sure this helps to explain anything.

[pedrocol]

One of the reasons why we see so much convection in the Amundsen sea when using the gade line (see third Figure below), may be the brine rejection that happens at surface due to sea-ice formation conditions at depth.

This paper propose a parameterization that distributes the brine rejection at depth within the mixed layer. "a power-law profile is used, with a greater fraction being deposited at its base than at the surface". "The main effect of the parameterization is to reduce significantly the mixed layer depth in ice-covered areas."

https://www.elic.ucl.ac.be/modx/users/fichefet/articles/2015_Barth%C3%A9lemy_et_al_OM.pdf

[pedrocol]

Following the previous comment, I think I may have at least a partial answer on why we have this behaviour in MOM5 and not in NEMO. NEMO by default produces frazil only at the surface, while in MOM5 this is optional. By setting frazil_only_in_surface = .true., we remove the extra negative heat flux that produces sea-ice at depth, avoiding the extra brine rejection but keeping the freezing point temperature at depth. I tested this for 3 years and the overflow in the Amundsen sea is reduced, the following plot is the age anomaly (frazil at surface true minus false) of a section in the Ross sea

The impact in the temperature is also noticeable

We see older and warmer water when frazil_only_in_surface = .true., which denotes the reduction of the undesired overflow. However, the on-shelf dynamics remain the same, and the entrainment still happens from the surface. We can see that in a section of the Amundsen sea, age anomaly with control

Is hard to tell with such a short run, if this will reverse at some point. What is certain, is that by setting frazil_only_in_surface = .true., we reduce the undesired overflow.

[pedrocol]

Brine rejection becomes dominant in those grid points where basal is injected. Using a brine rejection parameterization seems to solve this issue. In the following plot we can see basal+brine in the middle, and standard basal in the right, after 10 months.

The brine rejection parameterization consist in distributing brine down to the shelf front, in those points where basal is injected. It is also distributed down to 10m up to 2 grid points away from where basal is injected, in order to avoid "jumps" in the mixed layer contour.

The large convection causes cold basal waters to sink and reach the bottom, with the brine parameterization this undesired convection is drastically reduced.

[fabiobdias]

MLD in September (left column) and maximum MLD (right column) for the last 3 simulated years. There is not much difference among all the basal experiments in September, but significant ones in the maximum MLD. I think there is a clear improvement with the Brine scheme, but there are still considerable convection in the Amundsen on the BasalGadeBrine.

@pedrocol could you confirm the exact location of the Fig. 6 transect?