| 14 Jul 2026
Impact of the Tibetan Plateau and Rocky Mountains on deep ocean circulation during the Middle Miocene
Abstract. The stages of deep ocean circulation during the Miocene remain uncertain, but palaeogeographic changes likely played a role in its evolution to the Atlantic-dominated overturning state of today. Regarding orographic changes, the uplift of the Tibetan Plateau is thought to be conducive to a shift from Pacific to Atlantic deep water formation, with flattening of the Rocky Mountains having the opposite effect. Here, we perform a suite of fully coupled climate model simulations using the Community Earth System Model version 1.2 (CESM1.2) with middle Miocene palaeogeography to assess the impact of orographic and CO2 forcing on the meridional overturning circulation (MOC). The control simulation exhibits a strong Pacific MOC (PMOC; ~20 Sv) and no Atlantic MOC (AMOC). Flattening or widening the Tibetan Plateau has no significant effect on either basin's overturning. Flattening the Rocky Mountains strengthens the PMOC by 4 Sv by altering freshwater routing into the Arctic, but does not affect AMOC. A lower CO2 simulation (2x PI CO2) yields no qualitative change in overturning structure, indicating that the PMOC is not solely a consequence of warm climate forcing. Our results suggest that Miocene orography influenced Pacific deep water formation but was insufficient to trigger a shift toward modern-like AMOC conditions. This implies that other tectonic and ocean gateway changes were likely necessary to enable the development of the modern Atlantic-dominated overturning regime.
In their manuscript "Impact of the Tibetan Plateau and Rocky Mountains on deep ocean circulation during the Middle Miocene" Naik et al. explore the impact of different forcings on the strength of meridional overturning circulation in the Atlantic basin (AMOC), Pacific basin (PMOC), and in the Southern Ocean (SOMOC) during a time period when meridional ocean circulation was very different from today, which is here largely related to differences in Earth's geography. Based on CESM1.2 climate simulations with a Miocene geographic baseline, the authors consider differences in carbon dioxide forcing on large-scale ocean circulation and focus on Rocky Mountains and Tibetan Plateau as potential triggers for differences in three different components of the meridional overturning circulation (MOC). One of the key results is, in my opinion, the relative insensitivity of both AMOC and PMOC to the implemented geographic changes. The authors commend further research to reconcile this behavior with observations from other models in the upcoming MioMIP2.
I have read the manuscript with great interest. Generally the work is written up very well. The research topic is certainly relevant, a better understanding of the impact of geographic changes during the Miocene will be useful for the scientific community. Presentation of the research is timely with the next round of MioMIP. I find the manuscript generally suitable for publication in Climate of the Past. Nevertheless, I also think that there is still potential for improvement of presentation of the research and regarding evaluation and interpretation of the results in the light of previous work. Please find my detailed comments below.
Major comments:
Methodology:
In my opinion, the methodology needs more clarity. The various simulations employed here stem from two different simulation sets that have been presented in two different manuscripts: this work and the previous study by Renoult et al. (2026). At times I got really confused which simulations originate from which of the two ensembles, and which specific forcing and boundary conditions are actually linked to them. Adding a new Table 1 that clearly outlines for ALL simulations that feature anywhere in this manuscript, including those that overlap with (or ar are inherited from) the work by Renoult et al. (2026), would be beneficial. Please clearly describe the settings made for albedo, CO2, geography, and also add a note that highlights provenance of a simulation (Renoult et al. (2026), or this study). Simulation naming should be more consistent to avoid confusion. I found it difficult to derive, for example, which simulation has which albedo settings. I interpret that Mio_2xCO2 should have the same albedo as MioDS_Ctrl (but see my note below) - if so, then why not name the former simulation MioDS_2xCO2? If Mio_2xCO2 DOES NOT have the same dark soil, then I do not agree that the statement regarding insensitivity of MOC to different albedo settings made in lines 119ff must necessarily hold. Maybe more clarity regarding the different simulation characteristics and clearer separation of the impacts of CO2 and albedo would help here. Furthermore, please explain how the adaptation of topography has been done. In Figure 1 you show red squares outlining the regions of interest. At the example of the Tibetan Plateau flat simulation: has the whole region been set to 200 m? Or only regions that are higher than 200 m? Did you apply any further regional refinements? When increasing the size of the Tibetan Plateau, is this done via an anomaly procedure or absolute? Any offsets to be considered? Consider providing any relevant rules employed if applicable. This would help improving the documentation of your methodology. In the section experimental design one could also reflect on the potential relevance of hysteresis for your results, and in which way this was / was not considered here. This point could then be taken up again later in the discussion.
In this context I note that the statement "Given that these simulations are identical to Mio_Ctrl except for the bare soil colour" establishes that MioDS_Ctrl and Mio_2xCO2 each differ from Mio_Ctrl in the prescribed soil albedo. Yet, as far as I understand the manuscript, the authors leave open the option that all three simulations, Mio_Ctrl, Mio_2xCO2 and MioDS_Ctrl, could have soil albedo different from each other - equality of soil albedo between the latter two could be assumed, but is not a necessity. To avoid confusion and misinterpretation, please state explicitly, ideally in a table alongside CO2 concentrations and employed topography, which soil-albedo is used by each of the simulations.
Separately, the change in soil albedo relative to Renoult et al. (2026) is not motivated anywhere in your manuscript - neither the scientific reason nor its implementation details and amplitude of the difference. The previous publication by Renoult et al. (2026) has shown that soil albedo has a pronounced effect on the simulated climate if compared to other forcings. Your statement of negligible influence of albedo on the MOC could be more expressively reconciled with that previous finding, and differences of albedo in your simulation wrt. to that used by Renoult et al. (2026) should, in my opinion, be clearly motivated and quantitatively documented.
Presentation of results:
Generally very good, but see my comments below.
Using of a more strict metric to diagnose equilibrium of the AMOC is appreciated. I have the feeling that the related results could be documented in more detail. For Table 1 one could provide details on the fit model employed and provide more metrics than just the trends alone (e.g. R², p value, etc.)
Regarding presenting the mechanisms at play in your model, you show wind fields and their changes (Fig. 6) - this, in fact, has an influence, but is only one contribution to the water vapor transport that may be like-wise important (and that you do not show, while referencing it in your manuscript in line 266f) - adding an additional panel showing the change in water vapor transport may thus be useful.
Interpretation of results:
Different drivers that impact MOC: In your manuscript you clearly focus on the "top-forcing" of the AMOC, i.e. on anything that impacts on the ocean's density structure via changes at the upper ocean. This is an important contribution. Yet, towards evaluating the total response of the AMOC to forcing changes, it would be helpful to consider also other drivers, like wind stress and upwelling. Influence and behavior of the latter should, if not explicitly analyzed in your manuscript, at least be more broadly discussed. I think the study by Baker et al. (2025) provides useful context. The authors study AMOC stability in a warming climate and find a link between AMOC and PMOC via Southern Ocean upwelling. They furthermore illustrate that in the near future an AMOC could be maintained just by the fact that the development of a PMOC is too weak to compensate for reduced AMOC downwelling. In other words, a surface forcing that suppresses AMOC could be mitigated by an upwelling budget constraint. Your study looks at this problem from the opposite perspective and under a different topographic setting, but one could speculate that a similar relationship might also hold under these conditions. This would be something to reflect on in your discussion. Did you look at Southern Ocean upwelling and the related wind stress budget across the various simulations? If there were only minor changes, or if Southern Ocean upwelling was from the start too insensitive to the topographic changes implemented, then this could be leading to the hypothesis that PMOC is so strong due to the Southern Ocean upwelling budget being influenced too weakly, potentially because it is generally comparably strong in your model setup. In other words, is the AMOC maybe too weak because the PMOC is too strong? (You hint to that.) And is all this maybe at least partly driven from the south? (I do not think you explicitly raise this point yet.)
In a similar vein, findings could be impacted by the state and trajectory of the deep ocean. Are you sure that the model has been sufficiently equilibrated to exclude an impact on the derived results? In fact, you apply novel metrics of the AMOC to that end (de Boer et al., 2025), and this criterion appears to me stronger than what is commonly exercised. So, you already apply a quite strict criterion. Yet, there is previous work that suggests that simulations performed with the model family CESM1.2 belongs to, are sensitive to the initial state at the start of the simulation and to the spin-up trajectory itself. For example, Lee et al. (2025) have found based on iCESM1.2 that for the Miocene there may be multiple equilibria of the AMOC in dependence of the initial state. Curtis and Fedorov (2024a) show with CESM-1.0.4 that recovery of the AMOC after an initial suppression may take, depending on the carbon dioxide forcing strength, between 2,000 to 10,000 years, far beyond the time period when radiative equilibrium is reached, which is fostered by very slow temperature and salinity drift in the ocean. Curtis et al. (2024b) show, based on CESM1, that after a carbon-dioxide forcing-induced drop of the AMOC it takes ~2,500 years for a PMOC to arise, and it takes considerable additional time for the Atlantic-Pacific "seesaw" to lead to a reduced PMOC and an again recovering AMOC, all influenced by slow adjustment processes in the deep ocean that work on multimillenial time-scales. In fairness, given the computational demand, it is difficult to impossible to prove or refute any of these mechanisms with this model and the boundary conditions and forcings at hand. Nevertheless, my feeling is that this point should be more prominently discussed, and more early in the manuscript flagged, than currently the case (this is line 330ff if I am not mistaken). As far as I understand your arguments, at the moment you do not yet reflect on the strength of the PMOC being potentially controlled by the initial condition. Instead, you state that your model may be flawed by a salinity bias, herewith focusing on the ocean-surface forced response (and projecting your finding on a potential model error (line 330f), that may, or may not, be there), rather than reflecting whether the cause may at least partly be dependent also on the trajectory of the model state from its initial condition, which would not necessitate assumption of a model error. Are there arguments that would exclude the possibility of the impact of initial conditions and slow adjustment processes on your findings? If so, then I would clearly express them. If not, then, despite the very low trends that you observe, you may still be in a transient phase, and the results that you see on the AMOC and PMOC may therefore not be the final answer yet. In that case, a prominent note on potential incomplete equilibration would be appreciated, unless you can fully refute this possibility.
Â
Minor comments:
line 34ff: I suggest to also shortly reflect here on the upwelling route of the AMOC and its potential drivers as far as they are relevant for the study at hand
line 94: add "the" before "Parallel Ocean Program"
line 101: Can you provide any citation re. the Getech Plc. reconstruction?
line 104: change "were initiated" to "were initialized"? You speak here of ocean state initialization I assume.
line 106: What is the link of MioDS_Ctrl between both studies and why does albedo differ?
Fig. 1: I suggest to flip the order of the colorbar label to Depth / Elevation for consistency with the color order. Some more punctuation would help with readability of the caption (after (TP_wide), (TP_flat), and (Mio_Ctrl), I suppose). Per your caption statement one could add Mio_2xCO2 to the subheading of subfigure a.
line 115ff: I got quite confused here, and I think the link to results and simulations from previous work, and their combination with your own simulation ensemble, should be better illustrated. Your chosen CO2 range (3xPI) is in fact outside the ranges that you cite. Please consider providing an additional table as outlined above to clarify whether MioDS_Ctrl and the 2xCO2 simulation (Mio_2xCO2?) have the same albedo. Format of reference "Consortium" should be fixed. Regarding your last statement in lines 125ff, that soil excerts a negligible influence, where is this explicitly shown in your manuscript? Doing so could help interpreting your results.
line 126: You speak of negligible influence on the large-scale overturning circulation. Could there be also different effects that compensate each other?
Fig. 2: The plots are quite noisy (as expected). Are these already running means? For better visibility of individual simulation's behavior one could show the currently presented data with semitransparency and put on top a more intensly colored mean with a (wider) averaging window, reducing shown variability pronouncedly. Alteranatively or additionally one could show here also the regression lines that produce the results of Table 1. The latter maybe in a second figure showing the last 1000 years only and zooming in as much as possible on the relevant MOC range. I would increase font size at least for the legend.
line 144: remove the "and" before RM_flat?
line 156: The region for maximum AMOC is certainly adapted to the Miocene ocean state considered here. One could refer to relevant figures to indicate this motivation.
line 158: Using the plural (minima) suggests that there are multiple locations where you diagnose stream function - is this the case? And if so, how is the index computed?
line 159: equilibration-criteria? (for context)
line 170: I do not understand the formulation that ascerts that the PMOC was "arrested" - should this read "restricted" or similar?
line 170ff: Similar to other comments made by me, here one could also reflect on the potential impact of the inital state of the ocean that is the same for all simulations, and whether the Earth-system historical order of events analyzed here in parallel to each other could have any influence on the presented findings. I also think that this passage is more an interpretation than a result, so maybe move this text to the discussion section?
Fig. 3: please see my comments below regarding font size and potential for restructuring towards increased plot size made for Fig. 5
line 197: the statement "is not unexpected" could be reflected on regarding the various potential additional influences of the PMOC in the simulations shown here, see my other comments.
line 193: Has significance been tested? Otherwise I suggest reformulation.
Fig. 4: Please see my comments below (Fig. 5) regarding font size and potential for restructuring towards increased plot size.
line 223 (and 225): readability could be improved by making wider dashes and adding a space - at first sight these could be interpreted as hyphens connecting two words. Alternatively, split this very long sentence.
line 227: Is the term "theories" suitable in this context? Not sure, please check.
Fig. 5: I find plots and fonts very small. Maybe it would help to split the figure, showing the upper two plots separately as a reference, and making the remainder 4x4 panel a landscape figure. Regarding colorbars (this may well be personal preference) I remark that blue colors for reduced fresh water flux is at least to me counter-intuitive and confusing at first sight. Regarding runoff: I find the equation in the caption quite long, maybe move to main text and refer to it here. Furthermore, one could outline (somewhere in the manuscript, and reference this here) what components the runoff term includes, and how the model treats it. Is this the sum of all continental net-precipitation? Is it rerouted to the ocean by directly sending it to the nearest coastal point? Is it redistributed water shed wise, or does it maybe even follow some kind of more detailed river network?`Are there any remaining budget terms that are globally redistributed to the ocean, e.g. net-precipitation over ice sheets? Such details may have an impact on some of the upper ocean forcing fields you expose the model to, both their global distribution and potentially also the routing of freshwater to deepwater formation regions.
line 250: Here you state that the Tibetan Plateau has only local effects, which you explain by a minor change in elevation wrt. the simulation where this orographic structure is absent. Taking into account that the Tibetan plateau is the largest single body of orography in the region, and that, as you reference, it has been hypothesized before that the Asian summer monsoon system may be impacted by such an orographic change, it may be illustrative to more clearly demonstrate the magnitude of changes of the monsoon system that you find in your simulation, and to hypothesize why these changes may be insufficient to explain a more relevant role for the AMOC-PMOC system.
Fig. 6: I find text labels in many cases difficult to read, please increase font size. There is an overlap of figure subtitles with the vector scale - could be solved by defining in the caption an abbreviation for Surface wind and using this in subtitles instead. The caption should define the characteristic of surface winds shown (at 10 m height?). line 256 should probalby read "representing predominantly western winds ... predominantly eastern winds" - most vectors shown also have a meridional component. line 257: you describe "the contour", but I do not see any contours on the panels. Do you mean here color shading?
line 268ff: here you describe a mechanism that is disputed in the literature; since this mechanism is relevant to your work I suggest to provide more details that could explain different behavior found by different previous authors, and highlight, in how far this is potentially relevant for your study.
line 289: here you state that the co2 concentration is at the upper end of the reconstructed range. Yet, in fact, it is beyond that range (400-800 ppm) according to literature that you cite (Rae et al., 2021; Steinthorsdottir et al., 2025, your lines 73ff and 116ff). I suggest to rephrase for consistency.
line 287ff: here one could also refer to initialization state or hystersis of the AMOC, see my comments above
line 312: To clarify context of this sentence I suggest to replace "it is" by "PMOC".
References:
Baker, J.A., Bell, M.J., Jackson, L.C. et al. Continued Atlantic overturning circulation even under climate extremes. Nature 638, 987–994 (2025). https://doi.org/10.1038/s41586-024-08544-0
Curtis, P.E., Fedorov, A.V. Collapse and slow recovery of the Atlantic Meridional Overturning Circulation (AMOC) under abrupt greenhouse gas forcing. Clim Dyn 62, 5949–5970 (2024a). https://doi.org/10.1007/s00382-024-07185-3
Curtis, P. E., and A. V. Fedorov, 2024b: Spontaneous Activation of the Pacific Meridional Overturning Circulation (PMOC) in Long-Term Ocean Response to Greenhouse Forcing. J. Climate, 37, 1551–1565, https://doi.org/10.1175/JCLI-D-23-0393.1.
de Boer, A. M., Krishnan, S., Burls, N. J., Hutchinson, D. K., & Renoult, M. (2025). Evaluation of quasi-equilibrium criteria for coupled climate model simulations. Geophysical Research Letters, 52, e2025GL117040. https://doi.org/10.1029/2025GL117040
Lee, D., Sarr, A.-C., Acosta, R. P., & Poulsen, C. J. (2025). Multiple ocean equilibria and decoupling of Miocene atmospheric pCO2 and regional temperatures. Paleoceanography and Paleoclimatology, 40, e2025PA005126. https://doi.org/10.1029/2025PA005126Â
Rae, J. W. B., Zhang, Y. G., Liu, X., Foster, G. L., Stoll, H. M., & Whiteford, R. D. M. (2021). Atmospheric CO2 over the
Past 66 Million Years from Marine Archives. Annual Review of Earth and Planetary Sciences, 49(1), 609-641.
https://doi.org/10.1146/annurev-earth-082420-063026
Renoult, M., de Boer, A., Berntell, E., and Naik, T. J.: Shaping the mid-Miocene warmth: a sensitivity study on paleogeography, CO2 and model physics, Clim. Past, 22, 1203–1222, https://doi.org/10.5194/cp-22-1203-2026, 2026.
Steinthorsdottir, M., Montañez, I. P., Royer, D. L., Mills, B. J. W., & Hönisch, B. (2025). Phanerozoic atmospheric CO2
reconstructed with proxies and models: Current understanding and future directions. In A. Anbar & D. Weis (Eds.),
Treatise on Geochemistry (Third edition) (pp. 467-492). Elsevier. https://doi.org/https://doi.org/10.1016/B978-0-
323-99762-1.00074-7