Summary:

The manuscript describes an operational forecast model for the Baltic Sea and illustrates the applicability of the model by comparing a hindcast simulation to observational data from various sources.

Major comments:

The manuscript is well structured and the model seems well suited for the task at hand. I have only some minor points.  

(1) The purpose of the model could be stated more clearly - which variables are of interest for the end-users and forecasted? In this context I would find it nice to motivate the model assessment metrics and the choice of the considered observations based on the model purpose (why are these of interest and which precision of the forecast is required for potential end-users?)

(2) The assessment of the simulated sea ice is limited to a time series of sea ice extent. Some more assessment (including a visual impression of the spatial distribution) would be nice because I regard this aspect as very important for potential end-users.

More specific comments are listed below. Note that I am not an English native speaker either.

Specific Comments:

Line 1: The authors could add one sentence which model output they want to provide to the end user.

Line 7: I would suggest to add that the comparison is based on a hindcast simulation and the considered time period.

Line 17: Please add Dietze et al. (2014) (already referenced later)

Line 50ff: For my feeling the part about spurious mixing is a bit long in the introduction and parts of it could be moved to the final discussion when referring to the representation of inflows from the North Sea.

Line 64ff: The following parts seems a bit unrelated to the foregoing text. Maybe the authors could add a few sentences why operational forecasting is required in the Baltic and then come to the operational models. 

Line 70: better “….ocean circulation model….”

Line 80ff: I would suggest to add what should be forecasted/is of interest to potential end users.

Line 86: Instead of “skill metrics” I would rather use "model assessment metrics” - especially since the model is evaluated on a hindcast simulation and no real forecasts are considered. Note that this expression occurs rather often in the text and should be changed consistently.

Line 90: I find this confusing. Better? “..is an updated version of Nemo-Nordic 1.0 based on...”

Line 93: horizontal resolution

Line 93ff: I would finish the description of the model grid before coming to the boundaries. Also, it should be specified which open boundaries are used.

Line 99: was -> were

Line 102: How was the bathymetry modified (what counts as “shallow”)?

Line 106: How was the tuning done?

Line 135ff: I would suggest to finalize the description of the ocean part before describing the sea ice.

Line 173: I would rather call this section “Boundary conditions”

Line 174ff: The initial conditions could be described in a bit in more detail (e.g. from which conditions did the spin-up start?) because I expect at least some impact on the representation of the simulated deep water properties. Please note in this context that I see no perfect solution for their choice because the Baltic Sea virtually never reaches some kind of steady-state.

Line 178: Better? “The atmospheric boundary conditions are provided by ...”. Also, I would be interested in the spatial resolution of the atmospheric forcing.

Line 182/183: Better move to line 180 (+ delete “in NEMO”)

Line 190ff: Better? Observational data were provided by ….

Line 208: Better? “Model assessment metrics”

Line 231: I am not sure what is meant by “datum” in this context.

Line 235ff: The following part sounds like a lengthy excuse why SSH might not be captured perfectly in the model - which for me would rather refer to the discussion of the model results. Also, I would suggest to express this in a more positive way, i.e. what to expect from a SSH forecast. For my feeling the most important aspect for end users are deviations from the mean.

Line 252: When mentioning the locations (Kiel Holtenau etc.) the authors could refer to Fig.1c.

Line 259: This reads a bit confusing for me. Suggestion: The agreement between model and the ssh observations is generally higher in the open Baltic Sea than in Danish Straits. In the open Baltic the NCRMSD is generally below 0.3 and correlations between model and observations are above 0.95 (exceptions are....). In the Danish Straits, stations Frediericia and Copenhagen show much lower correlations of ...., respectively. This local drop in the correlations is expected due to the complex bathymetry in the Danish Straits.

Line 263ff: The BSH provides 2D maps for SST. It might be nice to show an example snapshot – even though the statistical value is of course limited.

Line 265ff: Maybe add "..mixing in the water column" because the other processes mentioned here refer to the atmosphere.

Line 265/266: ?

Line 281: The authors could add that they refer now to one of the ferries (for those who are not so familiar with ferry box data and the respective shipping lines). At which depths do the ferries measure?

Line 288/289: Which gradient?

Line 294ff: I would expect that the deep water properties are still impacted by the initial conditions because Döös et al. (2007) report a residence times of almost 30 years for the Baltic Sea. This can in my eyes not be avoided here but could be mentioned (unless this issue was investigated somehow and ruled out). In this section I personally would include an assessment of the simulated mixed layer depths.

Line 299: I would use some other wording instead of “Temperature profile skill” (similar expressions are used several times.)

Line 357ff: I regard the representation of sea ice as very important in a forecast model. I would find it nice to see a/some 2-dimensional map(s).

Line 392ff: For my feeling the discussion could focus more on the applicability of the model for forecasts. I like the honest discussion about MBIs.

Line 444: Why would a forecast model require a coupling with an atmospheric model? A coupled ocean-atmosphere model is not suitable for a 1:1 model-data comparison.

Line 445: Code availability: Pease provide doi and the link. I would find it nice if also the extra information to run the model was provided in addition to the code (such as initial conditions (restart), short test forcing/boundary conditions and specific model settings).

Reference

Döös, K., & Engqvist, A. (2007). Assessment of water exchange between a discharge region and the open sea–a comparison of different methodological concepts. Estuarine, Coastal and Shelf Science, 74(4), 709-721.

Revision of Nemo-Nordic 2.0 by Kärnä et al.

Summary:

The authors present and evaluate Nemo-Nordic 2.0, an operational marine forecast model of the Baltic Sea, which is based on Nemo-Nordic 1.0. The evaluation is sound and covers the most important aspects of  the Baltic Sea physical oceanography, and the paper has overall  a good structure and is in general well written.  This is an important paper that documents the development of the Nemo-Nordic configuration, and deserves being published after revision. 

Major comments:

-The introduction is not very coherent and needs to be revised. Maybe you could divide it into two subsections, with one about the modelling, and one more detailed about the physical oceanography of the Baltic and the North Sea? Alternatively, you could shorten the part on physical oceanography and only keep the most important aspects.  At the moment there is a lot of information on the physical oceanography in there without references (lines 25-34), please add appropriate references that support your description if you want to keep this text. 

-The model covers the Baltic Sea and the North Sea, but you only evaluate its performance in the Baltic Sea, and thus only a part of the model. Please describe why you do this. Still, I think that it is important, and that would be of value, if you also evaluate the model in the North Sea. If you want to focus on the Baltic Sea only in the main manuscript, you could maybe put some figures in supplementary information?

-You write that in this paper you evaluate the Nemo-Nordic forecast system. But, doesn’t a forecast system also  include data assimilation and forecasts? (or hindcasts). Indeed, you call your simulation a hindcast, but you do not evaluate its ability to “predict” the past,i.e. for how long the model manage to reproduce observations if starting from initial conditions created with data assimilation.  I do not think that you need to do this in the paper, and that it would be a paper of its own. I think that this is just a question of adding some extra text in the discussion/introduction about this, and/or revising the choice of words.

-Wouldn’t it be interesting to show how Nemo-Nordic 2.0 performs in comparison to Nemo-Nordic 1.0? 

Minor comments:

• lines 4-5: the 1 nautical mile horizontal resolution is an update as well no? 
• lines 15-17: These models do not only simulate the circulation… maybe it is better to write: “several ocean circulation models have been set up for the Baltic Sea”, or something similar
• lines 84-85: you repeat “as well as” twice in one sentence, please revise
• In section 2 it is not very clear what settings that are updates since Nemo-Nordic 1.0, and what you have kept the same, please clarify this.
• line 110: do you resolve baroclinic eddies in your configuration?
• line 174: 14 months are not enough to spin up the deep Baltic Sea. What did you use to initialize the spin-up run?
• line 244, and also elsewhere in the manuscript: how do you define good, relatively good and other, similar, qualitative words?
• figure captions and in the text discussing these figures: please describe if it is based on  daily or monthly output. It should also be mentioned in the methods what time-frequency your analyses are based on. A bit in a similar manner as you do for the comparison with the Ferrybox data.
• Could you put some lines in figure 1 showing the routes of TransPaper and Finnmaid ferries? As it is now it is difficult to relate figures 6-7 to a geographic location.
• I like figure 8, but I have some questions related to it; Why did you choose the upper 10 m for the surface layer? The summer thermocline is generally located deeper. When showing the skill as you do for the surface and deep layer, it does not tell us if the bias is due an eventual mis-placement of the thermocline/halocline, or if it is the modelled temperature/salinity that is off. It would be valuable if you could evolve the text around this and discuss it, a bit like you do for the salinity at BY5.
• figure 9 and 10: please write that the difference shows the model-observations
• section 3.6: please write why you have chosen these specific locations for your analysis
• figure 12: describe in the caption what the blue lines show.
• from figure 9 it looks like the model is too diffuse in the vertical. Maybe you could add this to your discussion on lines 408-416.
• lines 429-430: do you have some references to other baltic sea operational models that you could put here?