Review for "The very-high resolution configuration of the EC-Earth global model for HighResMIP”

by Moreno-Chamarro, Arsouze, Acosta, Bretonnière, Castrillo, Ferrer, Frigola, Kuznetsova, Martin-Martinez, Ortega, and Palomas

This is my first review for this paper. The authors present the EC-Earth3P-VHR model configuration, a high-resolution global climate model developed for HighResMIP, featuring atmospheric resolution of about 16km and 8km oceanic resolution. The model shows improvements in key regions like the Gulf Stream and the Equator compared to lower resolution models, with reduced biases in some areas but increased biases in others, such as a larger warm bias over the Southern Ocean. The model also demonstrates better air-sea coupling in tropical regions. However, it tends to overestimate the oceanic influence on atmospheric variability at mid-latitudes. Overall, the EC-Earth3P-VHR configuration appears to offer enhanced opportunities to study climate variability and change on regional and local scales.

First of all, the paper is in my view well-written, understandable, and has basically no typos. The figures are all high quality and well done. A description of the EC-Earth configuration for HighResMIP is clearly within the scope of GMD.



My only minor comments are with respect to highlighting some key results better, and better embedding the study into previous work, also outside of Europe. I have provided some references below for that purpose that the authors can decide to include at their convenience, and also gave suggestions for potential additional figures that could make the study even stronger. Overall, the study in its present form is already very interesting, it lists key results that are encouraging for fellow coupled high-res modellers, and is worthy of prompt publication. I am providing line-by-line comments below that I’d suggest having included before the paper can be accepted.

#################################

Line-by-line comments:

l.59-60 I would suggest to cite relevant papers for these projects, for example Hoffmann et al. 2023 (https://doi.org/10.1016/j.cliser.2023.100394) for Destination Earth or Hohenegger et al. 2023 (https://doi.org/10.5194/gmd-16-779-2023 ) and Rackow et al. 2024 (already cited elsewhere in the study) for nextGEMS. There should be something from MetOffice for PRIMAVERA as well

l.64 have been “proven”

l.84 For single precision, there are other earlier examples, e.g. Váňa et al. 2017 (https://doi.org/10.1175/MWR-D-16-0228.1);

as Destination Earth and nextGEMS were listed, there is also Sarmany et al. 2024 (https://doi.org/10.1145/3659914.3659938) for IO considerations

l.91 Another extreme example next to Gutjahr et al. 2019 is AWI’s CMIP6 climate model (e.g. Rackow et al. 2019, https://doi.org/10.5194/gmd-12-2635-2019, Semmler et al. 2020 https://doi.org/10.1029/2019MS002009), where their ocean is locally finer than 10km and has been run with a 100km atmosphere. There might be other examples in CMIP

l. 104 Regarding “High-resolution modelling usually relies on single-model component”:



I think there are several examples for relatively high resolution in both components, e.g. Chang et al. 2020 (https://agupubs.onlinelibrary.wiley.com/doi/full/10.1029/2020MS002298), Small et al. 2014 https://opensky.ucar.edu/islandora/object/articles:14403, the high-res studies from the South Korean group (https://ibsclimate.org/research/ultra-high-resolution-climate-simulation-project/ and listed references there, e.g. https://www.science.org/doi/10.1126/sciadv.abd5109), and the DestinE, EERIE, PRIMAVERA and nextGEMS results of course as well

l. 137 “of the” -> “of”

l. 138 Is this OpenIFS or IFS?

l. 161 240 s and 720 s has large white spacing

l. 198 What are the novel source code changes?

l. 200 Which model workflow software?

l. 204/205 Is there a parallel version of this available now? Is this linked as part of your document?

l. 209 Can you write technical details of “the network” here? Otherwise this does not tell much

l. 279-281 This inconsistency could trigger a bigger adjustment potentially. From your experience, does this lead to a cooling or warming initially that gets levelled out during the spinup?

l. 296 “but for” -> “except for”

l. 310 “will enjoy” please rephrase

l. 314/315 This is a very encouraging result; in principle, this is what modelling groups have been hoping for to see with higher resolution. This point could maybe be highlighted, potentially with a figure, and be included in the abstract?

l. 389 delete “bias” in this line?

l. 401 40 deg N seems far away from polar influence, are you sure about this statement?

Figure 13: Hard to see anything on those maps. Could you try with other colors or try a different (shorter) range?

l. 459/460 The plus/minus refers to what, standard deviation of monthly values?

l. 493-495 This seems like another key result that is very encouraging and not covered with a separate figure.

l. 535 Sections -> sections

l. 561 The lack of ocean current feedback comes a bit as a surprise here and could be covered earlier in the model description as to how the coupling is implemented

l. 580/581 Maybe give another example here if you know it (e.g. US or South Korean references mentioned above if matching), but this statement might indeed be correct

l. 585 excessive whitespace after “performed”

l. 619/620 Another study in this direction would be Sein et al. 2017 (https://doi.org/10.1002/2017MS001099). The authors argue that resolution over the shelf areas northward of the Gulf Stream front is key, an area where the cold Labrador water from the north meets the warm Gulf Stream water.

l. 687/688 “and VHR does it faster and appears more stable after 100 years than HR and LR” Another key result, see above

Review of the paper by Eduardo Moreno-Chamarro et al. on "The Very-High-Resolution Configuration of the EC-Earth Global Model for HighResMIP"

The paper introduces and validates a high-resolution configuration of the EC-Earth model, utilizing the IFS T1279 atmosphere and NEMO ORCA12 ocean components. The validation confirms that the model is technically ready for use and, in many aspects, shows reduced biases compared to its lower-resolution predecessors, although some tuning is still required, particularly for the Southern and Arctic Ocean regions. The authors definitely did a great job. I therefore recommend this paper for acceptance after minor revisions, which I have summarized below.

My main concern is that the plots should be updated to meet the standards of the 202Xs. This is especially relevant for Figures 7 and 14. Based on visualizations I have seen on YouTube and in several presentations, the presented plots appear outdated.

L43: Although the message is clear I would reformulate this sentence which currently states that biases lead to something improved.

L155: Regarding the effective resolution for the ORCA grid, could you cite the appropriate reference or tell how you compute it?

L166: independent on which grid, atmospheric or ocean? In general which type of remapping between ocean and atmosphere do you use? Is it conservative? Do you conserve the global freshwater balance? If not, how large is the imbalance?

I am also curious if you couple the ocean velocities for computation of the wind stress as this could explain some of the biases. Also I would list the number fluxes and surface fields which are being exchanged.

Fig. 2 Does this include the I/O cost? How many cores were assigned to OIFS, and how many to NEMO? What would the scaling for the individual uncoupled components look like, or at what number of grid points per core would scaling saturate according to the technical reports of these two models? If I understand correctly, for a 50x50 distribution in NEMO, this would be (4322*3059)/(4800/2) > 5,500 grid points per core. Given that NEMO scales linearly until around ~500 grid points per core, the saturation is reached much earlier than expected, possibly due to coupling issues or I/O. Is this correct?

L303: You say here that the spin up for 50 years is insufficient but from the plot we clearly see that even 150 years are insufficient. In Fig. 3 (lower panel) I note that the warming drift in VHR is the smallest for the global ocean but it is the largest for the upper ocean (upper panel). Does this imply that the overly large warming drift in the upper ocean is compensated by a negative drift in the deep ocean? 

Regarding my previous comment, is there drift in global salinity or the system is conserved?

L342: While the precipitation bias over the ocean, especially the “dipole structure” bias, will be mixed by ocean eddies, likely resulting in a low ocean salinity bias, the precipitation bias over land, especially in northeastern South America, will lead to heavily reduced runoff from land, as seen in the Amazon River. I observe this bias in all three simulations. What is the origin of this bias? Is it also present in other CMIP-type models?

L408: I also observe large deep mixing which takes place in the Arctic Ocean north of the Bering Strait. Seems like the Arctic Ocean is well mixed there which shall not be the case according to my knowledge. Nevertheless the ice patterns look nice but you didn't provide the sea thickness. Could you augment Fig. 7 with sea ice thickness such that the story becomes more consistent. As you say thick ice and the excessive brine rejection are the likely reasons for what is happening.

L654: In my opinion, the negative aspects are also strongly reflected in the biases observed in the Arctic Ocean, which should be addressed in the discussion section.