Review of "The Mixed Layer Depth in the Ocean Model Intercomparison

Project (OMIP): Impact of Resolving Mesoscale Eddies" by Treguier et

al.

This is a well-written summary of the comparison of mixed layers in a

suite of OMIP2 simulations, including both coarse (1degree) and fine

grids. The results are provocative and provide

a benchmark and motivation for future work. I support publication and

only have a few comments.

--line 26: We never really "validate" climate models. Instead we

"evaluate" models.

--line 74: "period period"

--line 188: "of course" is subjective; suggest removing 

--In many many places in the manuscript, the word "resolution" is used

when you really mean "grid spacing". Resolution is a non-dimensional

number whereas grid spacing measures the distance between grid points, typiically in degrees or km.

In most places this quibble is not so important since we "know what is

meant" even if we do not say it clearly. But on line 425 one reads the

rather confusing sort of statement that can result when "resolution"

and "grid spacing" are interchanged: "with a horizontal resolution of

less than 1km". Does that phrase mean the grid spacing is coarser than

1km or finer than 1km?? This is the sort of confusion that a novice

(and experienced) reader can come across when "resolution" is used

when "grid spacing" is meant. I suggest clarifying all uses of

"resolution".

--line 575: The authors observe that the MLD is more widely spread

among the OMIP simulations than SST. This is a very important

statement that perhaps has been noted before but is worth emphasizing

more. It offers an important counter-point to those who discount

OMIP simulations for using a prescribed atmosphere and so "have the

answer given to them". There is a lot more physics going into the MLD

than just that provided by the SST and SSS.

--Now for my slightly nontrivial and somewhat self-serving comment.

Namely, the authors point to the sensitivity of the MLD to the density

threshold (Figure 2), the upper layer depth, and time sampling. In the end, they

propose a 10m upper layer starting point rather than the "top grid

cell" advised by Griffies et al (2016). I think this is a good

suggestion. Yet they also support the density threshold approach, even

though it suffers from the problems they note in their Figure 2, as

well as those problems noted by BM04 whereby the density threshold

should be a function of the SST/SSS given the nonlinear equation of

state.

These limitations and hyper-sensititivities motivated Reichl et al (2022) to

propose a potential energy-based threshold.

That approach also can use monthly mean T/S to directly compare model

to obs, and it can be implemented online. So it is a practical

approach and simpler than some methods like Holte and Talley, though

more complex than the density threshold method.

I do not ask the authors to add the energy approach to their analysis,

as that would be more work than I can reasonably request. But I do

suggest they be somewhat more circumspect about the density threshold

for future MIPs. I might be wrong, but at this point I think a

potential energy approach ala Reichl et al is a physically compelling

and practical approach that avoids many of the problems with the

density approach. 

END OF REVIEW