Review of "Towards high-fidelity simulations of coastal submesoscale baroclinic

instabilities with MPAS-O (vE3SM3.0.0) Part II: Realistic

experiments" by D. R. Schlichting, K. M. Smith, M. R. Petersen, 

R. D. Hetland, K. E. Hinson and D. Engwird

This manuscript assesses the ability of a global unstructured-grid ocean model with local mesh refinement to reproduce coastal submesoscale dynamics in the northern Gulf of Mexico.

Two MPAS-O configurations are evaluated (MPAS-O: JRA and MPAS-O: USGS), the latter being specifically designed to better represent the Mississippi-Atchafalaya river plumes in the northern Gulf of Mexico. These simulations are compared with two regional ROMS configurations of the Texas-Louisiana shelf, using KPP and k-omega vertical mixing parameterizations.

Following a detailed description of the model configurations and parameterizations, the authors show that both MPAS-O configurations are able to reproduce the vigorous submesoscale variability of the open Gulf of Mexico reasonably well. However, they fail to realistically reproduce the summer submesoscale eddy field over the Texas-Louisiana shelf. As a consequence, excessive freshwater remains trapped on the inner shelf throughout the year, leading to significant salinity biases relative to the ROMS simulations.

Several hypotheses are proposed to explain these discrepancies, including the sensitivity of the vertical mixing parameterization to tuning, the neglect of horizontal momentum associated with river runoff, or biases arising from bathymetric representation and topographic smoothing.

Overall, this study highlights the potential of unstructured-grid models for representing open-ocean submesoscale dynamics, while also identifying shortcomings that remain for accurately simulating shelf submesoscale processes.

General comments

This manuscript is clear, well written, and easy to follow. The figures are of high quality, and the abstract provides an accurate summary of the work. The paper presents a thorough assessment of the strengths and weaknesses of MPAS-O in representing submesoscale dynamics in the northern Gulf of Mexico compared with dedicated regional ROMS simulations.

My main concern is that the discussion of the causes of the MPAS-O biases, particularly the lack of submesoscale eddies over the shelf, remains somewhat limited.

For example, I was surprised to see that MPAS-O uses a first-order Forward Euler scheme for tracer time stepping. Such a low-order scheme is highly diffusive and may substantially damp tracer gradients and associated submesoscale eddy activity. It would therefore be valuable to assess whether numerical diffusion contributes to the weaker shelf variability. One possible approach would be to compare the effective horizontal resolution of MPAS-O and ROMS using kinetic energy spectra at selected locations, following the methodology of Soufflet et al. (2016).

Another important aspect that deserves more discussion is the representation of bathymetry using the z* vertical coordinate. This coordinate system may significantly affect shelf dynamics because current-topography interactions are generally less accurately represented than in terrain-following coordinate systems. A useful sensitivity test, even on a shorter simulation and perhaps with a coarser horizontal mesh, would be to substantially increase the number of z* levels (for example, to around 200) to evaluate the impact of vertical resolution on shelf circulation and submesoscale dynamics.

Similarly, bathymetric smoothing, commonly applied in terrain-following models such as ROMS, may also contribute to some of the differences between the simulations. This aspect deserves additional discussion and, if feasible, dedicated sensitivity experiments.

Overall, I believe this is a valuable contribution, and I recommend publication after minor revisions.

Specific comments

- Throughout the manuscript, a large number of acronyms are introduced (e.g., RYF, SYPD, etc.). Reducing the number of acronyms to only those that are essential would improve readability.

- Table 1: Thank you for providing this clear and detailed summary of the model configurations. I was particularly surprised to see that MPAS-O uses a first-order Forward Euler scheme for tracer time stepping, as this choice may have important implications for the representation of submesoscale dynamics.

- Line 113: "five years on 120 CPU cores" is this correct? Could this instead refer to 120 compute nodes?

- Line 157: I would expect the influence of sea ice on the representation of the Loop Current (LC) to be largely indirect, through large-scale circulation changes (e.g., the MOC), rather than through local processes. This point may be beyond the scope of the present study.

References

Soufflet, Y., P. Marchesiello, F. Lemarie, J. Jouanno, X. Capet, L. Debreu, R. Benshila, On effective resolution in ocean models, Ocean Modelling, 2016, 98, 36-50.