Dear editor,

Thanks you for forwarding me this paper to review.

In this paper, a biogeochemical model is developed for the North Pacific region and assessed as to its suitability as a tool to assist in fisheries management.

I want to thank the authors for an interesting study and what I believe will become a useful contribution to their field. This paper is generally a solid piece of work and well-suited for eventual publication but I have a few suggestions I would like the authors to consider before I recommend publishing.

My biggest concerns are about the data products used in model-data assessments. In a few instances they are using the same product for model initialisation, boundaries and model assessments (glory, tpxo, WOA), which is not an independent comparison.  I do see these sorts of non-independent comparisons as a useful tool as it shows how faithful the model downscales the original product. However, on their own and without more independent assessments, we don’t know whether the biases we are seeing is the model degrading  the initial/boundary products or if the model is actually improving on them.

My next but related concern is to do with the Aletutain islands. It seems likely to me that there are some very fine scale processes occurring around these that could have a noticeable effect at the 10-20km scales that you are assessing. If your model is a higher resolution than the data products then it is likely that your model is better capturing these processes than your data products so you will need some higher resolution data products to properly assess the model performance in this region.  I am not too familiar with the oceanography of the region, so it could be that the authors have already considered this - but I would like to see a short discussion on how well represented they expect this region around this island chain to be represented in their model and the observational datasets.

A few minor comments

There are quite a few acronyms which does reduce the readability, particularly in the abstract. I suggest carefully assessing which of these are really needed and which can be written out in full.

Fig 1 is on a different projection to the rest of the figures. I am less familiar with this region, and it did make it more difficult to locate the different regions on the subsequent figures.

lines 115-120. What sort of an effect does land-use changes have on this system?

Line 200: NEP domain -> NEP10km

Line 206: Suggest also including nudging timescales here.

Lines 305-306: It seems that for both of the mixed layer depth comparisons you are using the same calculation for model and observations (which is a +ve) – but I suggest rephrasing to make this clearer.

Line 345-350. It is hard to see where the 500 m contour is on fig 1, and it is hard to assess how many grid cells you have on the shelf. I suggest adding the 500 m contour to figure 1, and including some text to say how many grid cells wide this continental shelf region is (perhaps max and min lengths?) The shelf is quite narrow and for most of your domain, the area less than 500 m appears to be only a few grid cells wide. Is the 10km resolution model a good enough tool to describe the coastal region? If shelf conditions are a key fisheries-critical variable, then nesting into the coastal regional may be required (I am not suggesting you need to do this for the current publication). Your global observation products may also struggle to capture coastal processes. I think this paper could be complete without the coastal assessments, and a short discussion about how to approach this in future would work instead (for example, using in situ products that measure shelf scales combined with higher resolution models).

Line 394: What does CMEMS stand for?

Section 2.5.3: This section (and the similar results section) could be of value as it shows some of the good features that help your model code run faster. However, as written, I don’t think it is essential to be included as some of these results will be system specific (i.e. it is of interest to you, but not to a broader audience). If you include this section, then I would like to know more about where did your model runs. I note that you mention the computer at the end of the paper, but this needs to come earlier. I would also like to see a description of the computing system used as these results will likely vary across different computing architectures (e.g. inter-PE communication speeds will vary across different computers).

Line 428: How did you conclude that the 400s tracer time step was best?

Line 486: Can you indicate the Aleutian island chain on Figure 1?

Line 500 – 505:  I’m not convinced by this statement. It is possible that these biases represent your model improving on the tpxo dataset. In my experience (in other parts of the world), a locally produced model tends to compare better to tide gages than the tpxo model for partially enclosed areas. Is there tide gage data for this area that you can compare to? If resolution is to blame for the bias, how does the tpxo resolution compare to your model?

Figure 18: note here that some of the apparently poorer comparison on the shelf compared to offshore could also be because the offshore comparisons were log transformed.

Line 734 – Suggest writing CPA in full as it is not used often, and readers will have forgotten what it stands for.

Line 879: your fig 18 suggests that the coastal biases are not small

Line 968: Quite often when you increase resolution you also need to decrease timestep. This could potentially be a lot more than 8-fold increase in computational cost!