| 08 Jul 2025
The impact of model resolution on the North Atlantic response to anthropogenic aerosols
Abstract. A set of novel, idealised, single-forcing experiments were performed to isolate the impact of anthropogenic sulphur dioxide emissions on North Atlantic SST variability. The medium-resolution (60 km atmosphere, 0.25° ocean) and low-resolution (135 km atmosphere, 1° ocean) of the HadGEM3-GC3.1 model were used to investigate the impact of resolution on the forced response. The SST response at both resolutions is timescale dependent: a fast, large-scale surface cooling is followed by a slow, ocean-driven warming responses. Warming of the sub-polar North Atlantic is due to a strengthening of the Atlantic Meridional Overturning Circulation (AMOC) and is stronger at the medium resolution. This difference is related to surface density fluxes across the subpolar North Atlantic. The growth of Labrador Sea ice is stronger at low-resolution which inhibits air-sea interaction and reduces surface buoyancy forcing, leading to a weaker AMOC response. There is also evidence of a stronger AMOC positive feedback involving salt-advection at medium-resolution. These results show that the large-scale North Atlantic response to external forcing can be sensitive to regional differences, such as model climatology of Labrador Sea ice and its response to aerosol cooling.
Competing interests: Laura Wilcox is a member of the ACP editorial board.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
General comments
The authors present the results of two climate model runs of different resolutions under idealized frameworks of aerosol forcing. The main focus of the analysis seems to be on the North Atlantic climate response to this external forcing at multidecadal timescales, and under low to medium resolution model configurations. The article is generally well-written and structured. The conclusions are supported by the analyses presented and I believe the research is novel and of substantial value to the climate research community. However, there are minor corrections to multiple sentences and to the figures that should be addressed, and further corrections to the explanation of processes that are mentioned.
Overall, the figures are generally too small, therefore, I included a separate section with advice on how to improve them. The authors should re-write the abstract to reflect the specificities of their analyses. It seems as some topics addressed in different sections are forgotten in the summary and conclusions, including explaining why the Labrador Sea is so important in their simulations, the multi-decadal variability analysis, and near-surface heat fluxes’ imprint on net radiation and subsequent sea surface temperature.
Upon these corrections, I advise for publication.
Specific comments
Title: Advise to include the word “multidecadal” after North Atlantic.
Abstract: From the abstract the main objectives seem to be to isolate the impact of Sulphur Dioxide onto North Atlantic SST variability and the effect of model resolution on the AMOC. However, the authors do not mention the multidecadal variability (particularly AMV) that is so present in their introduction.
AMOC in density space vs. depth space: the analysis is centered already on the AMOC in density space, which is being recognized as beneficial to understand higher latitude processes. Perhaps the authors could leave the plots of the AMOC in depth space in the appendix to avoid confusion, and use the already detailed explanation as to why AMOC in density space was chosen.
Multiple instances:
Lines 4-5: “large-scale cooling is followed by a slow…”. What is the slow referring to here? the authors could add the simulation years to show the timescale, since it is not clear.
Lines 6-7: What happens to the density fluxes? Add an adjective to indicate how they are modified, and with respect to what – either the stronger AMOC or the higher resolution.
Line 7: “The growth of Labrador Sea ice” – you refer to the sea ice in the Labrador Sea? If yes, replace with “sea-ice growth in the Labrador Sea is stronger”.
Line 8: the surface buoyancy forcing here is always towards deep water formation, if yes, then it would lead to weaker AMOC. However, the buoyancy forcing can be either towards lighter or denser waters. Specify.
Line 17: Replace “decades of” with “at decadal timescales”.
Line 26: What is your historical period? 1850-2014?
Lines 44-46: I do not see the link between “Furthermore, the modulation of the radiative budget is not the only way AA can drive SST changes;” and “turbulent heat flux changes can be driven by the advection of cold, dry air from the continents instead (Robson et al., 2022)”.
Line 47: “Many of the beforementioned studies” and are these historical simulations based on CMIP6? Indicate what you mean by historical simulations.
Lines 49-51: This sentence needs to be clarified as it now only implies one modelling issue.
Line 54: “North Atlantic variability “change” in the coming decades”.
Lines 56-57: How is using MM and LL relevant to this so-called second question?
Lines 59-60: Replace model/models with version/versions. MM and LL are the versions of the same model.
Lines 62-66: As a personal taste, a paragraph with the article structure is not needed.
Line 69: “forms part” ?; and define CMIP.
Lines 75-79: Reformulate the paragraph. These are not model configurations but model components, and all of them are global, because that is your domain. Advice: “HadGEM3-GC3.1 is comprised of the GA/GL7.1 (Walters et al., 2019) atmosphere and GO6 (Storkey et al., 2018) ocean model components, coupled via OASIS-MCT coupler (Valcke et al., 2015), in timesteps 1 hour. Additionally, GA/GL7.1 implements the GLOMAP modal aerosol scheme, whereas the ocean component, based on version 3.6 of the Nucleus for European Modelling of the Ocean (NEMO) ocean model (Gurvan et al. 2017) also embeds the sea ice model GSI8.1 (Ridley et al., 2018).”
Line 81: Personal taste – remove “in CMIP6 nomenclature”.
Line 82: This resolution is only in the mid-latitude atmosphere?
Line 90: Which parameter?
Line 92: Is ORCA025 eddy-resolving? Therefore, there is no GM parameterization? If yes, reorganize sentence to specify that GM is not applied in ORCA025 because of its resolution.
Lines 108-109: This sentence should be somehow included in the abstract.
Line 114: “teleconnection mechanisms” – like?
Line 128: Changes in?
Line 131: Personal taste – remove “Lastly, for the PDRMIP simulations, the ocean outputs were not archived.” – this information is not relevant to this study.
Line 132: It seems like this subsection of the methodology is part of the results section.
Line 141: “indicates upward anomalies in Figure 2”
Line 143: Move “(3-7 days, Rasch et al. (2000))” to after “sulphate aerosols”.
Line 159: “and subsequent AMOC strength” instead of “and also an indicator for how strong the AMOC is likely to be”.
Line 160: Careful with causality here. The sentence implies that AMOC is directly modulated by TOA imbalance due to aerosol forcing, which is not the case. Perhaps restructure sentence to indicate to which extent it occurs.
Line 165: “0.25Wm-2” – superscript.
Lines 195-196: “stronger or weaker” instead of “a speeding up or slowing down”
Line 198: Reformulate “Net DWSW is also sensitive to surface albedo changes, such as from changes in the sea ice.” to detail the causality.
Line 200: How did you find out that THF and net DWSW explains more the SHF?
Lines 202-214: What is the physical meaning of the colors in your figures? A negative anomaly in surface heat fluxes indicates more heat loss from the ocean to the atmosphere, or from the atmosphere to the ocean? What is the direction of these fluxes?
Line 204: What is your domain for subpolar North Atlantic?
Section 2.3: How does the previous section connect directly into this section? This is missing.
Line 216: To me it does not seem as if different sections of the North Atlantic respond differently to anything, but that surface fluxes are heterogeneously influenced by aerosol forcing in the North Atlantic.
Line 219: Which changes? You haven’t analysed the ocean heat content but ocean heat fluxes at the surface.
Line 221: “even as SO2 emissions abruptly decrease after simulation year 30”.
Line 226: replace “stronger warm” with “warmer”. Additionally, the warmer anomalies are concentrated in the subtropical to subpolar North Atlantic. It is important to define your domain of subpolar North Atlantic.
Line 228: To which sigma is your AMOC in density space referenced?
Line 246: In what sense does the medium resolution have a stronger AMOC response?
Line 267: W/m2 – consistency. Before you used Wm-2, even. So, it should be Wm-2.
Line 303: It would be interesting to see the rest of the years in the appendix.
Lines 304: this has already been explained in lines 296-298.
Line 314: Have you analysed interior mixing? See Sidorenko et al., 2020, 2021 for the role of interior mixing in the AMOC.
Line 318: extremely dense with respect to what? Also, from the plots it seems that the marginal seas connected to the Atlantic are the ones with more outcropping. Not the Arctic.
Lines 323-326: Perhaps this conclusion changes if you look at density fluxes at the isopycnal of stronger AMOC, instead of at the surface.
Line 339: either AMOC-sigma or AMOC-. Consistency is important here.
Line 345: density fluxes towards where? They have directions. Same in line 350 and other instances addressing fluxes.
Line 358: “Sea ice growth inhibits”. Additionally, it does not inhibit air sea interactions but heat loss to the atmosphere. From figure 3 it also seems that there is a pronounced cooling in LL where sea ice grows more, not the opposite.
Lines 370-371: Again, inhibits heat loss, not air-sea interactions.
Line 380: By surface forcing you mean heat forcing? To which changes and anomalies you refer to? AMOC strengthening, weakening, deepening, shoaling? Lighter, denser waters?
Line 385: “… Arctic. Jungclaus …” instead of “… Arctic, Jungclaus …”.
Lines 386-388: Personal taste. The last sentence can be less inquisitive and show that this is an outstanding question.
Lines 392-397: The description of thermal and haline terms can be combined for conciseness. Both are similar but with one held constant, so having them described separately is redundant.
Lines 397-399: Why do you calculate the relative importance this way? What do you mean by square? Square root, T2? Whatever it is also needs to be in the title (i.e. T2-S).
Lines 409-411: “In the western and central SPNA, which includes the LS and IS, both models show positive density anomalies, although with different relative contributions from temperature and salinity.”
Line 418: “From 1500-3000 m”.
Line 433: Ocean advection of what?
Line 444: Move “resolutions” to after the parenthesis.
Lines 447-466: This is conclusions, should be at the end.
Line 448: SPNA is in the Northern Hemisphere. How can the Northern Hemisphere be lagged to itself?
Line 449: “which is more pronounced” instead of “with a stronger AMOC response”
Line 451: Double parenthesis.
Lines 455-458: Which differences? What do you mean by mean state here? Also rephrase the analyses of sea ice growth. It is a little odd in the sentence.
Line 473: Interrogate?
Lines 476-478: It is the same model but tuned for different resolutions
Line 486: It is unlikely so. Usage of large ensembles have their advantages. And your model is not necessarily better, or your setup is necessarily better. It is different. Careful with overselling the results.
Line 501: You add more insight on how high resolution can also imprint its own bias due to the stochastic nature of climatic processes. Perhaps look at Hirschi et al. 2020.
Figures
Most figures could be plotted as a subplot to decrease the space between each subplot and improve representation, as figures are too small in general. To that, colorbars and axes which are shared should be omitted. In terms of timeseries, if the years are not dates (i.e. 1850-1900), it is simulation years. Additionally, some figures are plotted focused only on the North Atlantic, while others also show the global domain. This should be consistent unless there is a specific global analysis, which is not the case for SSTs, for example, when the authors only describe changes in the North Atlantic but plot it for the entire globe. Another common feature is the repetition of information in figure titles that is not necessary as they are already described in the caption.
Below is the specific description of suggested changes for each figure.
FIGURE 1:
FIGURE 2:
FIGURE 3: Compare processes rather than resolutions. This way one can easily see the difference between MM SST and LL SST one beside the other and the colorbars can be bigger and shared. Put the years the left as y-axis label. Figure will become bigger and differences will be easier to see.
FIGURES 4-11: Same as suggested for figures 1-3.
FIGURE 8: Why show the haline component in the figure if it is not in the same order of magnitude? Clearly the salinity contribution towards denser or lighter waters is not relevant here. I advise to plot the density flux at the isopycnal of maximum AMOC strength, there the haline component might play a role, instead of at the surface. Otherwise, the plot can be only of the total density flux, since it is just influenced by the thermal component.
FIGURE 9: Consistency with MM and LL instead of N126 and N96. This is related to resolution but you have not used this convention in the methodology.
FIGURE 10: No need for MM-LL in the title.
FIGURE 11-12: No need to include the depth interval. This can be in the figure description.
FIGURE 13: define -ve, +ve and so on. This figure is also quite convoluted. Perhaps be more specific as to which anomalies you refer (cooling, warming, salinization, freshening, and so on).