The report of Schiller et al., entitled 'Atmospheric Forcing as a driver for Ocean Forecasting' illustrates different ways of provisioning surface momentum, heat, and freshwater fluxes to operational ocean models.

The report presents two kinds of flux dataset sources (observational vs. numerical prediction systems) and then gives some relevant considerations about the application of atmospheric forcing to ocean forecasting systems for global vs. regional/coastal systems. The brief conclusion does not give any recommendations, except that the suitable atmospheric forcing depends on “the applications and users”.

I understand that the report can not be exhaustive about the available atmospheric or flux products and can not list in details how each operational ocean forecasting system is currently driven in surface. Nevertheless, the report is here more confusing that clarifying, especially section 1.

Here are in details my main concerns:

- Using observations is, by definition, a way to drive an ocean monitoring system or to produce a (re)analysis. Obviously, using an atmospheric forecast appears mandatory to do an ocean forecast. Somehow, this is never clearly mentioned in the paper.

- To my knowledge, surface fluxes are not directly observed by remote-sensors, but are computed using different geophysical observed variables, generally from different platforms, and using parametrization for computation. It could be interesting here to mention if there is any initiative to gather and evaluate specific satellite flux (or atmospheric near-surface parameters) products designed for operational oceanography.

- For ocean forecasts, the use of an atmospheric forecast as surface forcing can be done by 4 methods:

• using directly the atmospheric fluxes produced by NWP systems of weather services/centres. For that, the relevant questions for OOFS are the data availability, space-time resolution and domains for regional/coastal OOFS;
• using a so-called “bulk” forcing, i.e. the near-surface atmospheric parameters. This method permits to use the ocean surface explicit variables (temperature, current, albedo) to compute inline and eventually at each time step the turbulent fluxes and the upward radiative fluxes, and so to introduce a pseudo-coupling. This method brings the same questions than the first one, plus, the choice of the surface flux parametrization that is here crucial;
• using an intermediate simplified atmospheric model (e.g. ABL1D) driven for the large-scale by the atmospheric NWP 3D fields and producing surface fluxes consistent with the ocean evolution and resolution;
• a full ocean-3D atmosphere coupling but with specific issues relative to the numerical cost and the initialisation/assimilation, but the advantages (compared to the 3 first methods) i) to have no (or for regional OOFS a lower) dependence to the data availability from external providers and ii) to ensure a two-way consistency.

In my opinion, an improved way to present information about atmospheric forcing for OOFS can be done by following the suggested outlines hereafter:

1. atmospheric forcing for ocean forecasts. There come only NWP systems as possible forcing, but with the methods and considerations explained before, and additionally the issue of open boundaries/surface forcing consistency for regional OOFS, that is well described in the current section 2.2.
2. atmospheric forcing for ocean analyses/monitoring systems. There could be a discussion of using atmospheric analyses or “observational” flux products;
3. atmospheric forcing for re-analyses/OOFS evaluation/past case studies. For this purpose, using reanalyses or any best fit of observed data is clearly recommended.

With these comments and suggestions, I recommend a revision of the paper.

This paper briefly introduces the state of the art of atmospheric forcing into ocean model forecasting, with a focus on the air-sea flux datasets.

General comments:

Although I understand this is an introductory chapter, I miss a brief discussion on the main limitations of current satellite-derived flux products in terms of the spatial and temporal resolution required for ocean forcing. Also relevant, NWP output is not only used because of its low latency but also (and mostly) because of its ubiquity. Moreover, some more emphasis on the current limitations of the NWP output in terms of its relatively poor spatial resolution and quality in the ocean forcing context is also desirable.

Regarding satellite-derived flux datasets, I believe too much (positive) attention is given to SAR-derived wind stress in the context of coastal forcing, while no explanation on the current limitations of such technique is provided. Although quite some efforts have been devoted to SAR wind retrievals over the past two decades (see publications from, e.g., Horstmann, Mouche, Grieco, Moiseev, Zecchetto, Zhu, etc.), there is currently not a single SAR wind processor that can provide a coastal wind stress product of sufficient quality and/or coverage for use in operations, while its use for OOFS development purposes must be done with caution and on a test-case basis.

Specific comments:

1. L47: Wind stress is well-determined from scatterometers since SEASAT-A (1978) and ERS-1 (1991). Suggested references: Jones et al. (1982), Stoffelen and Anderson (1997), Portabella and Stoffelen (2009).
2. L48-50: Similar to the CMEMS L2 OCN product, Khan et al. (2023) use a very old technique (Portabella et al., 2002) to systematically derive coastal wind vectors from SAR. Many publications (incl. Portabella et al., 2002) point out the limitations of such technique, in particular the lack of small-scale variance in the derived wind direction component (which is mostly driven by the background wind direction, i.e., the NWP wind direction). Moreover, the uncertainty in the wind direction component is then propagated into the wind speed retrieval. I would therefore not recommend the use of this product (or actually any other SAR-derived wind stress product to date) for coastal ocean forcing purposes and use it with caution for OOFS development purposes.
3. L65-66: “…has the potential to produce biases, particularly in the radiative flux fields and precipitation (Trenberth et al., 2009; Weller et al., 2022) and in the wind stress vector components (Belmonte and Stoffelen, 2019; Trindade et al., 2020)”.
4. L67: Please, explain why atmospheric reanalyses are suitable for OOFS development.
5. L82-83: Please provide references. Also, briefly explain how do satellite data supplement NWP output in forcing ocean models. For example, Trindade et al. (2020) show how scatterometer-derived wind stress can be used to remove NWP model output local biases.
6. L116: Please, name a few high-resolution regional NWP models and add corresponding references.
7. L125-130: Please, moderate the benefits of using SAR-derived wind (stress) for coastal OOFS development purposes.

References:

Belmonte Rivas, M. and Stoffelen, A.: Characterizing ERA-Interim and ERA5 surface wind biases using ASCAT, Ocean Sci., 15, 831–852, https://doi.org/10.5194/os-15-831-2019, 2019.

Jones, W. L., Schroeder, L. C., Boggs, D. H., Bracalente, E. M., Brown, R. A., Dome, G. J., … & Wentz, F. J. (1982). The SEASAT-A satellite scatterometer: the geophysical evaluation of remotely sensed wind vectors over the ocean. Journal of Geophysical Research: Oceans, 87(C5), 3297-3317. https://doi.org/10.1029/jc087ic05p03297.

Portabella, M., and Stoffelen, A., “On scatterometer ocean stress,” J. Atm. and Ocean Techn., 26 (2), pp. 368–382, https://doi.org/10.1175/2008JTECHO578.1, 2009.

Stoffelen, A., and Anderson, D., “Scatterometer data interpretation: derivation of the transfer function CMOD-4,” J. Geophys. Res., vol. 102, no. C3, pp. 5767-5780, 1997.

Trindade, A., Portabella, M., Stoffelen, A., Lin, W., and Verhoef, A., “ERAstar: a high resolution ocean forcing product”, IEEE Trans. Geosci. Rem. Sens., 58 (2), pp. 1337-1347, https://doi.org/10.1109/TGRS.2019.2946019, 2020.