The study focuses on the development of a regional ocean physics-biogeochemistry model for accurate downscaling of the ocean conditions in the North-West Atlantic; with main application/focus on providing information for supporting marine resources management. The performance of the model is evaluated based on a hindcast simulation during 1993-2019. The comparisons mainly focus on: (i) large-scale and model-wide physical/biogeochemical conditions near the surface and the Gulf Stream position, (ii) regional features of interest for ecosystem variability.

The study is interesting, well-written and aims to provide a much-needed efficient tool for simulating and projecting ocean conditions on regional scales along the US east coast. The model configuration, simulations and the methods for the model validation are well-described, including many details as appropriate for a study focusing on model development. The choices of the “metrics” for the model performance are well justified and are relevant to the dominant circulation, and ocean conditions and ecosystem variability in the region. The scope fits well the “Geoscientific Model Development”. I want to highlight that the manuscript is excellent in its current form. However, I was a little disappointed that there were no comparisons in terms of the vertical structure of the water column and water masses representation (beyond the Northeast channel at intermediate depths). I also have some other minor queries as I describe below. Hence, I recommend the following minor revisions/clarifications before the study gets published (mainly to offer the opportunity to improve the manuscript).

General comments:         

1. Water masses – Vertical sections. The comparison metrics used are very informative. However, in my opinion, it will be helpful to add some comparisons between the model and observations/reanalysis in terms of key vertical sections (or profiles) to better demonstrate the representation of different water masses in the model. For example, I suggest adding such comparisons for the winter and summer temperature and salinity climatologies for either: (i) sections of depth versus distance form the coast to beyond the shelf break, for maybe the Gulf of Maine+Georges Bank+slope, the MAB+shelf break and the Gulf of Mexico; or (ii) representative vertical profiles (perhaps averages) over the 4 EPUs. Consider doing the same for phosphate (or nitrate) and perhaps DIC.

2. Main circulation features. In my opinion, it would beneficial to have a visualisation of the model mean circulation. I suggest adding a figure/map that compares the surface (or the vertical averaged) currents in the model to the reanalysis. The authors can decide how best to visualise this (perhaps using arrow plots with a background colormap for speed).

3. Streamlining (this is just a suggestion): The introduction is very informative; however, in my opinion, the discussion for the ocean conditions and variability in the North West Atlantic (lines 24-90) is a little too descriptive/long and could be condensed to better highlight the shortcomings due to limited availability of skilful high-resolution regional predictions and projections. I suggest streamlining lines 24-90, but this is just a suggestion and up to the authors’ preference.

4. Figures and latitude longitude (this is just a suggestion): I understand that the map-figures do not have latitude and longitude so as to preserve space and make them look more compact. However, adding latitude and longitude to all of the map-figures would help readers who are not familiar with the region to easily follow the features shown in these figures; particularly in figures that involve zooms in specific regions (e.g., Figures 6,13, 14, 20, 21, 23).

5. Seasonal mean estimates. The seasons for temperature, and chlorophyll (Figures 3, 12, 23) are defined as: (i) December-February, (ii) March-May, (iii) June-August, (iv) September-November; while the seasons for nutrients (Figures 10 and 11) are defined as: (i) January-March, (ii) April-June, (iii) July-September, (iv) October-December. I was wondering if there is a reason for using different months to define the seasons for nutrients. If so, please explain it in the text. If not, it might be better to keep the definition of the seasons consistent for all the fields.

Specific comments:

6. Line 133. I am not sure what “… coastwide extend to address the prominent cross-boundary issue expected under climate change” means here. Maybe it refers to along-shelf propagating signals, such that for example you have a large domain covering the whole US East coast? If yes, in my opinion, your domain is still affected by cross-boundary issues, as the Labrador current is not resolved in the domain but rather prescribed at the north ocean boundary. Please consider clarifying/re-writing. (This is not a criticism as all the regional models are subject to this, but more a request for clarification.)

7. Figure 1. Consider adding the names of some key regions in figure 1.a: e.g., the Northeast Channel, Cape Hatteras, Texas, Louisiana, Florida, Gulf of Mexico, Gulf of Saint Lawrence etc..

8. Lines 209-213. There is an emphasis on obtaining reliable solutions without applying restoring (e.g., surface salinity restoring). In my understanding, the simulation covers about 25-30 years (+ spin-up), and I was wondering if 25-30 years is a short time period in terms of the model developing a significant drift. Hence, I am not sure if the absence of restoring in the 25-hindcast run is indicative of the model’s performance in longer simulations (e.g., climate projections), in terms of the emergence of significant drift; and if it will actually be beneficial to include a strategy for accounting for this drift in regional climate projections with your model. Maybe I have misunderstood something, but I suggest adding a brief discussion to clarify why it is expected that there will be no need for restoring to account for any drift in your regional ocean-only model under long-term simulations (longer than 25-30 years).

9. Lines 228-229. Please can you clarify how the rivers runoff salinity and temperature is treated (e.g., are you prescribing/assuming a constant 0 PSU salinity for river runoffs, or observed values?).

10. Line 404-405: I am not sure I understand what the feedback of biogeochemistry to tides will be in you model? (So I am not sure why there is expected to be even a negligible feedback). Please, can you clarify what this small feedback would involve (at least to the reply, as I was confused).

11. Lines 590-592 and Figure 7. To me, based on the 0.4 m shift in the colormap, it appears that the absolute dynamic topography and the model SSH have a difference/bias in magnitude. Is this maybe associated with the estimates of absolute dynamic topography and the geoid (I am not an expert on this so I am just curious about it)? I suggest, for clarity, to add the difference between model and observed SSH and the equivalent metrics as in the other figures (bias, RMSE, MedAR and Corr). This would help to better understand the magnitude and significance of the difference between the two datasets.

12. Figure 12. I suggest that you add the difference between model and OC-CCI satellite in the figure, as it is a little difficult to compare by eye where the model overestimates or underestimates the surface chlorophyll.

13. Figure 13. If I understood correctly, this is a zoom-in of figure 12 (maybe mention this in the Figure 13 caption). I suggest adding the bias, RMSE, MedAR and Corr metrics in the figure for the two different regions.

14. Lines 727-728, Figure 21: It is a bit difficult to compare by eye the observations and model sea-ice concentration in Figure 21. Please consider adding the difference between observed and model in a third panel, as well as the associated metrics (bias, RMSE, MedAR and Corr).

15. Lines 731-733. Please consider highlighting in the text that this is for the Gulf of St Lawrence, for example “ … in sea ice coverage in the Gulf of St Lawrence, with correlation …).

16. Typo, Lines 87-89: decadal time scale is repeated, maybe consider removing the “At decadal timescales” in the beginning.

2023.09.18

Review of gmd-2023-99:

 A high-resolution physical-biogeochemical model for marine resource applications in the Northwest Atlantic (MOM6-COBALT-NWA12 v1.0),

by AC Ross et al.

This paper describes a 1993-2019 simulation of a 1/12 degree MOM6 ocean model including the COBALT BGCM. The model output is compared to a suite of physical and biogeochemical data for the time period, and the study includes an interesting section on computational performance. The paper is well written with the various analyses and results clearly explained. I have no major objections, with just a few minor points to respond to (see below). I consider that the manuscript requires only minor revisions to be acceptable for publication.

A general comment: The study presents an extensive model-data comparison and in this reviewer’s opinion the model truly does do a good job of simulating most variables and time series. However, I do not know of any set of quantitative skill metrics which have an agreed upon level below which a model is deemed not fit for purpose. So there is no real way to judge whether a model passes the test. I note that the authors comment briefly on this in the Conclusions section (L865) with a comment about an acceptable level of correlation between model and data.  Perhaps I am wrong about this lack of generally accepted skill level(s) but I would be interested in the authors’ thoughts about this.

Comments:

1. L31: The 2018 Brickman et al. paper documents and reveals the mechanism for these intrusions and should be referenced here:
   1. Brickman, D., Hebert, D. and Wang, Z., 2018. Mechanism for the recent ocean warming events on the Scotian Shelf of eastern Canada. Continental Shelf Research, 156, pp.11-22. https://doi.org/10.1016/j.csr.2018.01.001
2. L73: Regarding NAO effects on fisheries. The Fisher et al. 2008 paper provides an excellent example of the NAO effect on fish distributions in the model area. The authors may want to check the paper out and include a reference to it.
   1. Fisher, J.A., Frank, K.T., Petrie, B., Leggett, W.C. and Shackell, N.L., 2008. Temporal dynamics within a contemporary latitudinal diversity gradient. Ecology letters, 11(9), pp.883-897.
3. L158 …163: Does the z* coord system have partial bottom cells? Authors should clarify the vertical resolution of the bottom cell.
4. S2.3 Spinup and Hindcast: L332-342: I am not sure that I followed the spinup procedure, or perhaps I do not understand the logic. The authors describe a 10y spinup for the BGCM component (using a perpetual 1993? -- clarify) which is then used as the initial BGCM field for the main model run, which starts from rest in 1993 using the Glorys TS field. This confuses me. Because the BGCM is part of “main model” then the physics model must also be spun up. Why not use the 10y spinup to start the physics model as well? It is rare to not spinup a model, even if it is initialized from a 3D “spun up” TS field. Please clarify this procedure.
5. L364: conservatively interpolated; some details on this would be helpful
6. L384: “introduce a small bias”; further explanation needed.
7. L385 … (and Fig 6), re GS position: There are a number of recent papers discussing changes in the GS position. For a slightly different analysis the authors should have a look at Wang et al. (2022) [Wang, Z., Yang, J., Johnson, C. and DeTracey, B., 2022. Changes in Deep Ocean Contribute to a “See‐Sawing” Gulf Stream Path. Geophysical Research Letters, 49(21), p.e2022GL100937. https://doi.org/10.1029/2022GL100937]
8. L453, re EPUs (Figures 1b, 18): For researchers working on the SS/GSL/NL region, the SS EPU would not be considered part of the SS. This should be changed to eastern GoM (EGOM) as in Pontavice et al.
9. Results: 3.1
   1. GS position (F6): F6a,b: lon/lat on these panels please
   2. L593: “cross-shore” is a bit confusing; consider north-south or meridional
   3. No need for Fig12. Fig13 shows the results better
   4. L646: for clarity add the word “model” before “mesozooplankton”.
10. S3.2
    1. Deep Salinity: Fig19: (a) TS fig. Please make colored symbols in the legend bigger so they can be seen. (b,c) plots of %water masses (LSW & GSW): No mention of corrs in caption and it is not clear that the Glorys12 values are the model vs Glorys12 or Glorys12 vs one of the data series
    2. Ice: F21,22: many models have problems with the advance, extent, and retreat of sea ice in the GSL (and on the NL shelf). This model does very well in this regard.
11. S3.3 Computational performance: Authors are commended on including this section. A couple of questions:
    1. L771 (last sentence): Re scaling efficiency: The reason is not clear to me as wouldn’t the (15x15 point / PE) restriction from the 2D BT solver still apply even when the BGCM component is implemented because the physics module still uses the BT solver? Please clarify.
    2. L783: I do not understand this sentence, in particular the statement about reduced number of PEs needed by the longer thermodynamics timestep. I would have thought that the longer timestep requires fewer clock cycles, not fewer PEs. Perhaps my confusion is related to my comment above(?). In any case please clarify this
12. Discussion
    1. L810: The Brickman et al. (2018) reference is relevant here as well.
13. Conclusions:
    1. L866: I note that the authors mention a 0.5 corr as “commonly considered lower bound for useful prediction skill” although they do not provide a reference