REVIEW of

" Assessment of the Finite VolumE Sea Ice Ocean Model

(FESOM2.0), Part II: Partial bottom cells, embedded sea ice and

vertical mixing library CVMIX” by                         Scholz et al.,  2021.

The presented manuscript describes and evaluates the application of the recently developed   unstructured-mesh Finite-volumE Sea ice-Ocean Model version 2.0 (FESOM2.0).  The current state of the art FESOM2.0 model include multiple options and the authors analyze the impact of the  partial cell, embedded sea ice-ocean model coupling and several mixing parameterizations from the Community Vertical Mixing (CVMIX) project library on the  performance of the coarse resolution global sea ice FESOM2.0 model.

 During the last decades the FESOM model was actively developed and applied for sea ice modeling in multiple publications. Therefore, the detailed analysis of different FESOM2.0 options is a necessary step and will be extremely useful for potential FESOM users. The manuscript is relatively well written and provides many Figures which illustrate the impact of different options on the model solution. Meanwhile, I find that the authors provide only “technical and qualitative comparison” and that makes this manuscript to be very similar to a technical report but not a research paper.  

Thus, I definitely may recommend the manuscript for publication in the Geoscientific Model Development Journal, but recommend adding some quantitative comparison between different FESOM2.0 options and provide at least a qualitative physical explanation of the revealed differences. This is especially important towards  understanding the impact of the partial cells and embedded sea ice options.  

Line 54-55   - “Embedded Sea ice … “

As I remember, the first understanding of the importance of the  Embedded Sea ice was provided by Hibler et al., 1998,  and after that, was used in several publications (e.g. Hutchings et al, 2005). I guess Hibler’s embedded ice models are different from FESOM2.0, but these publications are also related to the embedded  sea ice and should be at least cited.

Line 57   “implementation of embedded sea ice relies on the zstar vertical-coordinate option in FESOM2 and also on  the fact that the sea ice component is called on each time step of the ocean model”

As I understand FESOM2 uses an EVP solver (or its modification) and this suggests the application of two time steps. Which time steps do you mean for the ocean and sea ice models? Also, the modern sea ice model (e.g. CICE6, MIT ocean model)  typically includes explicit and implicit solvers: EVP, different VP solvers (GMRES, Newtonian..)? Do you plan to include an implicit VP solver? If so, it could  be expensive to use a VP solver for each time step of the ocean model.

Line 151   we limited the thickness of the partial bottom cell to be at least half of the full cell layer thickness to reduce the possibility of violating the  vertical Courant–Friedrichs–Lewy (CFL) criterion.

It looks like a strong limitation: usually the CFL is defined by dz near the surface, so it looks strange to not allow the near bottom dz to be about 0.25 from the full dz. Did you observe any instability in the FESOM numerical scheme?

Liner 159-184

In order to demonstrate the effect of the partial cells on the simulated ocean state we performed two model simulations using the full cell and partial cell approaches, respectively. We investigate, first, the temperature biases of the full cell approach with respect to the data of the World Ocean Atlas 2018 (WOA18, Locarnini et al., 2018; Zweng et al., 2018, in the left column of Fig. 1) and, second, the temperature differences between partial cell and full cell (partial-full) averaged over five different depth ranges 0-250m, 250-500m, 500-1000m, 1000-2000m and 2000-4000m (in the right column of Fig. 1)

When you compare the FESOM2 performance in this (or similar) figure, you provide only a qualitative comparison ,which sometimes looks subjective. You need to provide some quantitative criteria (e.g. STD from WOA ) for  each (0-250m, 250-500m, 500-1000m, 1000-2000m and 2000-4000m ) layer and include these numbers into the Figures. This will provide real estimates of the positive impact of the partial cells and/or other FESOM2 options.      

Lines: 168-174 + Figure 1

This is a very interesting figure, could you please explain the intensive zonal features in Full Cell cases? Note, sometime these features change the sign between 0-250m and 250-500m.

I see some “correlation” with figure 12, from Adcroft et al., 1997, so, some analysis would be very useful.

Figure 4

Could you please provide the reference related to the intensive convection in the Weddell Sea and south from Greenland?

Line 216-217 – “One can summarize that partial cells lead to a clear improvement of the circulation pattern, especially regarding the branch of the Gulf Stream and NAC even in rather coarse resolved configurations.”

Where can I see that?  Maybe anadditional plot?

Line 222    (linfs)

(linfs)-> LinFS?

Line 234   “…  reality sea ice and ocean velocities are rarely identical especially in the presence of high frequency wind” 

Not accurate: even if wind is 0, there is a turning angle in the ocean forcing, so, ice will be flowing in a different direction.  

Line 236;   variability especially near the ice-edge where ice divergence/convergence is large (Campin et al., 2008).

Are you talking about MIZ? In the miz, the concentration is low and ice is especially thin. Not sure that may provide significant impact into the overall sea ice dynamics.  

Line 247;  Figure 6f.

As I see, the embedded ice model provides less ice in the deep Beaufort Sea and more ice in the shallow ESS sea. Could you please provide any explanations?

The same for the Figure 6e – anomaly in the Greenland Sea in March. Why?

Lines 261- 263?  The temperature and

salinity differences reveal that a significant warming of up to 0.5°C and a salinification of up to 0.10 psu occurs in almost the entire AO due to embedded sea ice, except in a thin stripe along the eastern continental shelf of the AO that shows negative anomalies in the depth ranges of 0-250 m, 250-500 m and 500-1000 m

Question:  I always suggested that embedded sea ice  provides a stronger divergence/convergence due to stronger ocean forcing, and that  the embedded sea ice should  provide more “open” water or polynyas and that should cool the near surface layer. Am I wrong?

Did you estimate the “embedding” impact on the sea ice divergence/convergence?

Line 265-… : The changes in temperature and salinity can be explained by the changes in ocean currents.

Do I understand this correctly: you explain the temperature/salinity increase due to stronger inflow of the Atlantic water into the AO? If so, you need to estimate this inflow through the Fram strait and provide the numbers.  

Figure 7; -

You analyze the impact of the embedded sea ice in the AO and Southern Ocean. Why do you need the area between 60S and 60N? It is hard to see any features in the AO. Could you please re-plot this figure and remove the area 60S-60N and increase the AO /Antarctic region?

Line 393: “… all topographic features which is induced by the tidal vertical mixing parameterization”

Which topography features? What does it mean ALL? Besides, you may mention the increase of diffusivity mixing in the Indonesian region where internal waves are well-known features of the local dynamics.

Figure 14   ….

Something wrong with capture… Check it.

Figure 15

I see a strong increase in Weddell Sea in the Antarctica> ? Any evidence /references that tides are active in this area?

Lines 437-439- 

Discussion of figure 18 should be NOT be before the discussion of figure 17.!

Change the order of the figures.

Line 477 + Figure 17

Figure 17 should be Figure 18. See above.

For all figures:

1)How is it possible that the coastline for 2000-4000m is the same as for 0.50m?  Strange.

2) As I mentioned above,  the disruption of these figures require some numbers which will allow you to estimate the difference between different mixing and numerical schemes utilized in FESOM2.0.  It is reasonable to provide these quantitative differences for each layer.

3)  Physical explanations of the observed differences between different schemes/options is necessary, as well, in most of the cases.

References:

Hibler W., P. Heil, and V. I. Lytle, 1998: On simulating high-frequency variability in Antarctic sea-ice dynamics models. Ann. Glaciol., 27, 443–448

Hutchings et al, 2005, Modeling Linear Kinematic Features in Sea Ice, Mon. Wea Rev, 3481-3497, 2005.

The manuscript of Scholz et al documents a large number of sensitivity tests related to new features for FESOM2 (partial bottom cells, sea ice coupling, and vertical mixing).  I was extremely impressed by the vast number of simulated years done for this manuscript and the quality of FESOM developments and the writing.  While I found the paper sound and clear for the most part, I have a couple major concerns and more detailed minor comments that I would ask the authors to consider.

First and foremost, I struggled in the manuscript to see the scientific impact to the broader community.  This manuscript reads as a FESOM technical report and is no doubt very beneficial to that community.  However, it was very difficult for me to draw out points of interest to the broader modeling community.  A few examples, I felt the discussion and results around partial bottom cells (PBCs) and sea ice coupling did not come across as of interest to the broader community, but it is quite possible that having all these results in one place will be a useful go to reference.  You could consider trying to address causes of biases more directly and what the role of the change was in the circulation change, or perhaps consider making a recommendation of best practice configuration for FESOM at the very least.  These could be ways to improve the broader impact of this manuscript.  As of now, for the most part biases are simply noted and then moved on.  There are a few exceptions, it is mentioned the way FESOM computes the bulk Richardson number causes the changes in some biases, but plots of boundary layer depth or surface layer average velocity and buoyancy for each method were not plotted.

Second, I think there were issues with the discussion of vertical mixing with regards to CVMix.  It is stated that FESOM_kpp is configured in a different manner from CVMix, with the surface layer averaging noted as a key difference.  However, this is not correct.  CVMix leaves a number of choices up to the calling model, amongst them is the velocity and buoyancy difference for the bulk Richardson number.  Griffies et al 2015 recommends using 10% of the boundary layer depth as the surface layer average, but this is not within CVMix itself.  We further discuss the dependence on model choices in Van Roekel et al 2018. As an example, POP chooses the largest shear between a depth and the surface cell making this one step further than what FESOM chooses. It would be interesting if FESOM made other choices in configuration relative to default CVMix, e.g. Monin obukhov/Ekman limiters, matching at the boundary layer base, shape function parameters, etc… and what the impact of these choices might be. 

Relatedly, I think it is important to note that the FESOM KPP choice of the first layer being the surface layer is not physically consistent with KPP (even Large et al 94).  Throughout KPP there are built in assumptions regarding the depth of the surface layer (default to 10% of the boundary layer depth) and assuming the first layer as the surface layer is inconsistent.  While it is a valid assumption it is important to point out this issue and the consequences of it.  It would be interesting to explore the impact of this choice, but that is likely beyond the scope of this paper.  Here it shows basic plots of T/S, but would be good to know a finer scale view too.  Have you conducted a simulation that uses the CVMix library but the FESOM_KPP choice for the numerator and denominator of the bulk Richardson number?  This could clearly show differences associated with the choice.  My expectation is that the shear is the more dominant term and using the surface value will deepen boundary layers, but it would be interesting to see clear evidence of this.  Again, this is a possible place where you could make broader impacts. 

To help clarify and grasp the broad points of the paper, I would suggest perhaps having bullet point take summary somewhere (perhaps the discussion/conclusions) with call outs to key figures.  A few other suggestions to help with clarity: (1) a table of differences in KPP vs PP and in CVMix / FESOM versions would help maintain clarity (2) a table of the FESOM KPP configuration, e.g. is matching utilized?

Minor comments

Throughout, you write CVMIX, the acronym is CVMix.

In numerous places you have things like “southern hemisphere September” which reads odd to me.  I’d suggest parenthesis around the month.

There are also a number of references that need proofing, e.g. Griffis 2015 and Ilicack2006.

There are also many places with subscripts that didn’t typeset correctly.

Line specific Comments

L27 – Delete “The”

L29 – suggest adding “southern hemisphere” to “sea ice melt season mixing”

L40 and L41, remove “one” after “first” and “second”

L50-57 – by the word “embedded” I expected the sea ice code to be in the ocean code as in MOM6/SIS, is this the case for FESOM?  It is not clear.  It may be better to say “non-levitating” or “pressure exerting” for clarity, but no need to change the word if clearly defined if the ice code is in the ocean model or uses a coupler.

L57 – you mention that you must compute sea ice at every ocean time step to “embed”, this doesn’t seem desirable and is not actually required in our experience in MPAS/E3SM.  Embedding is more dependent on the fidelity of the ice-ocean coupling in our experience.  Can you clarify what you mean by this statement?

L81 – It isn’t clear to me what “prime vertical mixing” means, is this default?

L82 – the phrase “deliver a usable mixing scheme” is confusing to me.  Do you have a meaning in mind?

L129 – delete the comma and which

L145 – It is unclear to me why Pacanowski and Gnanadesikan 98 is discussed.  It seems FESOM uses Schepetekin 2003 instead.

L151-153 – Have you tested FESOM without the requirement that the bottom thickness be greater than ½ the layer thickness?  As an example, MPAS-O runs stably without this requirement, but I have not looked in depth at the possible biases that may exist even though it is stable.  It could be interesting to further examine this choice.

L157 – vertice -> vertex

L165 – biasin -> bias in

Section 3.2 – In this section you discuss a dependence on sea ice thickness in configuration choices but then only present comparison to sea ice concentration.   It would be helpful to plot thickness.  This is also more consistent with what is actually measured by satellite.

L281 – while MOM6 does have a branch with CVMix the original implementation was designed to reproduce the POP formulation, so I would change MOM6 to POP

L315+ - the strong similarity between KPP and PP in the analysis was surprising to me.  These schemes are quite different, especially in the near surface.  Have you examined boundary layer differences?  E.g. Mixed layer depth?  I wonder if perhaps the similarities are due to the fields presented, you show averages over fairly thick layers and below the boundary layer I imagine FESOM uses the shear instability induced mixing of Large et al. 1994, it is possible your analysis is only highlighting more deep ocean impacts and the similarities in the LMD SI induced mixing and PP81 mixing make the results seem similar.  A simple test would be comparison of MLD between the schemes.

L398 – ando -> and

L450-453 – Why not test different background options?  Seems like a very easy test to do.

L486-488 – this sentence is very confusing to me.  When you use ‘except’ but discuss freshening in one part and temperature in the other it doesn’t read easy to me.

L527-529 – have you tested combinations of changes?  it seems possible (perhaps likely) that some changes have nonlinear interactions and is not as simple as just adding biases.

L549-552 – any ideas why you see a large change in the gulf stream for the MOMIX + KPP?  Is this related to changes in transport (AMOC maybe?)

L582-583 – As an MPAS-O developer I confess I agree with your statement here, I’m always deeply impressed by the pace and quality of FESOM developments.

L701-702 – add commas around “and to a decrease in the high-latitude”

Fig 14 and 18 – the plot titles seem wrong in most panels here.  Also in panel (c ) of both there is an odd high salinity bias 40N.  It is interesting it is identical in both Fig 14 and 18.  Is this a plotting or analysis artifact?