Relative contributions of CO<sub>2</sub> and ice sheets on the Mid-Pleistocene Transition

Millot-Weil, Jeanne; Valdes, Paul; Farnsworth, Alexander

doi:10.5194/egusphere-2026-3719

Preprints

https://doi.org/10.5194/egusphere-2026-3719

Preprints

08 Jul 2026

| 08 Jul 2026

Relative contributions of CO₂ and ice sheets on the Mid-Pleistocene Transition

Jeanne Millot-Weil, Paul Valdes, and Alexander Farnsworth

Abstract. The Mid-Pleistocene Transition (MPT) represents one of the most prominent and debated climate reorganizations within the long-term cooling trend of the last 3.6 million years. Although the role of ice sheets, greenhouse gases and their combined effects have all been implicated, their individual contributions to the MPT remain unresolved.

Here, we assess their respective roles using a unique set of four series of fully coupled equilibrium palaeoclimate model simulations spanning the last 3.6 Ma at 4-ka intervals, constrained by realistic boundary conditions. These simulations differ in their prescribed insolation, greenhouse gases and ice sheet forcings where a realistic baseline time evolving scenario is analysed against a series where each forcing is kept static throughout to constrain their individual impact on the MPT.

First, we evaluate the performance of the four sets of simulations including the combination of the three changing forcings against geological data. We find that the model captures the main characteristics of temperature variations identified by geological records, such as the global cooling trend over the past 3.6 Ma, acceleration of cooling at the MPT, with amplification of the glacial-interglacial cycles along with a change of pace from a 40-ka to a 100-ka cyclicity. Then, we compare the four simulated timeseries to untangle the individual role of each forcing through time and show that greenhouse gases exert a direct and dominant role on both the long-term global cooling trend and the shift in glacial-interglacial cyclicity associated with the MPT. Ice sheets primarily modulate the amplitude of glacial-interglacial variability through their influence on sea-ice formation and ocean circulation.

Received: 24 Jun 2026 – Discussion started: 08 Jul 2026

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.

Download & links

Jeanne Millot-Weil, Paul Valdes, and Alexander Farnsworth

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-3719', Anonymous Referee #1, 04 Aug 2026

This manuscript presents a large ensemble of fully coupled climate model simulations to investigate the effects of greenhouse gases and ice sheets on Pliocene–Pleistocene temperature evolution and the Mid-Pleistocene Transition. The four sets of 919 snapshot simulations represent a substantial computational effort, and the resulting dataset has clear potential value. The results show that greenhouse gases play a direct and dominant role in both the long-term global cooling trend and the shift in glacial-interglacial cyclicity associated with the MPT. Ice sheets primarily modulate the amplitude of glacial-interglacial variability through their influence on sea-ice formation and ocean circulation. However, the current manuscript is dominated by statistical comparisons and does not yet provide a sufficiently detailed physical explanation of the simulated responses. My major concerns are as follows:
1. The central weakness of the manuscript is that its conclusions are based primarily on statistical contrasts among the simulated temperature time series. Trends, glacial–interglacial amplitudes and spectral characteristics are useful for describing the model response, but they do not identify the physical pathways through which greenhouse gases and ice sheets affect temperature.
2. The manuscript argues that ice sheets influence temperature indirectly through changes in sea ice and ocean circulation. This could be an important result, but the analyses presented do not yet establish these mechanisms. Moreover, if changes in ocean circulation are invoked as a central mechanism, the authors need to demonstrate that the 500-year snapshot simulations are sufficiently equilibrated.
3. The effects of greenhouse gases and ice sheets on temperature may depend on the background climate state. The manuscript would therefore benefit from comparisons among several representative climate states, including a warm Pliocene state and pre- and post-MPT glacial and interglacial states. For each state, the authors should examine not only the magnitude of the temperature response but also the underlying physical processes. Such analyses would help determine whether similar changes in greenhouse gases or ice sheets produce different temperature responses under different background conditions and would identify the mechanisms responsible for any state-dependent climate sensitivity.

Citation: https://doi.org/10.5194/egusphere-2026-3719-RC1
- AC1:
  'Reply on RC1', Jeanne Millot-Weil, 29 Sep 2026
  This manuscript presents a large ensemble of fully coupled climate model simulations to investigate the effects of greenhouse gases and ice sheets on Pliocene–Pleistocene temperature evolution and the Mid-Pleistocene Transition. The four sets of 919 snapshot simulations represent a substantial computational effort, and the resulting dataset has clear potential value. The results show that greenhouse gases play a direct and dominant role in both the long-term global cooling trend and the shift in glacial-interglacial cyclicity associated with the MPT. Ice sheets primarily modulate the amplitude of glacial-interglacial variability through their influence on sea-ice formation and ocean circulation. However, the current manuscript is dominated by statistical comparisons and does not yet provide a sufficiently detailed physical explanation of the simulated responses.
  Dear reviewer, thank you very much for your helpful remarks and suggestions.
  My major concerns are as follows:
  The central weakness of the manuscript is that its conclusions are based primarily on statistical contrasts among the simulated temperature time series. Trends, glacial–interglacial amplitudes and spectral characteristics are useful for describing the model response, but they do not identify the physical pathways through which greenhouse gases and ice sheets affect temperature.
  
  Thank you for sharing your reservations about this work, this is something we are also keenly aware of.
  Statistical analysis are useful and robust to characterise the temperature response to greenhouse gases, ice sheet and the combination of both, and how it evolves through time. They are also useful for comparison with geological reconstruction. Moreover, our set of snapshots are especially designed to address the evolution of the climate. Physical pathways through which greenhouse gases and ice sheets affect temperatures have been repeatedly highlighted in the literature notably through the work from Model Intercomparison Projects (CMIP, PMIP, PlioMIP…). These are generally done with just one or two snapshots. Concerning the MPT, the main unresolved questions about which physical pathways triggered the transition concern interaction between greenhouse gases and ice sheets (see Berends et al. 2021).
  As stated on lines 468-472, since greenhouse gases and ice sheets are imposed forcings in our simulations, they do not interact with one another, and we cannot resolve this question. Instead, what we propose to do in this work is to quantify the impact of greenhouse gases and ice sheet changes on temperature trend and variability. Future work will focus on underlying mechanisms triggered by changing ice sheet and greenhouse gases, however this requires a new modelling framework to do so. This is already being planned and we hope in the near future we can tackle this aspect robustly.
  As suggested by Reviewer 2, the title and abstract were misleading regarding the nature of the results presented in this paper. We have therefore revised them to better reflect the scope and interpretation of our results.
  The manuscript argues that ice sheets influence temperature indirectly through changes in sea ice and ocean circulation. This could be an important result, but the analyses presented do not yet establish these mechanisms. Moreover, if changes in ocean circulation are invoked as a central mechanism, the authors need to demonstrate that the 500-year snapshot simulations are sufficiently equilibrated.
  
  Thank you for these important comments.
  We agree that with the runs may not all be in equilibrium with only a 500 model year run. However, our experience from previous work is that 500 years is reasonable compromise. For instance, the original snapshot simulations in Singarayer and Valdes (2010) used 500 year snapshot simulations for the last glacial cycle. Subsequent work extended these to more than 3000 years yet the surface climatology was relatively little changed (except for the discovery of some oscillations at 30kyr, see Armstrong et al. 2022).
  In order to try and minimise drifts, we initialised the latest runs to the nearest equivalents from above. This was very effective for the more recent periods but was poorer worked effectively for the last 2 million years. For the older time periods, they were mainly initialised from modern.
  We can also diagnose whether there are any significant trends in the model runs. Top of the atmosphere energy imbalance are all within 0.3W/m2, and trends in the surface temperature are less than 0.05C/century (which was the PMIP2/3 spin-up criteria). Deep ocean temperatures do have some trends, but they are relatively small. 90% of the simulations have volume integrated temperature trends less than 0.05C/century. The largest trend is 0.08C/century. Of course, in practice it is impossible to exclude the possibility that these trends could produce larger surface changes due to thresholds being exceeded but we feel we have good evidence that our surface climates are robust.
  Concerning sea ice, some of our analysis has shown that Northern hemisphere sea ice may contribute to shaping the 100-kyr cyclicity shift in the OrbIce Northern hemisphere air surface temperature (NHΔST). More precisely, the 100-kyr cyclicity shift disappears from the wavelet power spectrum of OrbIce NHΔST when grid boxes covered by sea ice are excluded from the calculation (see Fig. A3 in the initial manuscript). Since the wavelet power spectrum of OrbIce Northern hemisphere sea surface temperature (NHΔSST) (see Fig. 10b in the initial manuscript) does not show this shift either, we inferred that sea ice contributes to the emergence of the 100-kyr cyclicity shift in NHΔST in OrbIce. Nevertheless, we agree that this cannot be used as a proof for a general indirect influence of ice sheet through sea ice. We have therefore reformulated the parts of the manuscript that are misleading on that subject (notably the abstract).
  Concerning the impact on ocean circulation, this was indeed not highlighted in the initial manuscript. The fact that OrbIce Southern hemisphere sea surface temperature (SHΔSST) exhibits the 100-kyr cyclicity shift during the MPT, unlike OrbIce NHΔSST, while the OrbIce Southern hemisphere air surface temperature (SHΔST) signal does not appear to be strongly affected by sea ice or ice-sheet extent calculation (see Fig. A3 in the initial manuscript). This led us to investigate the potential role of ocean circulation in shaping Southern hemisphere temperature variability.
  A 500-year simulation alone is indeed not sufficient to fully equilibrate the deep ocean. However, we observe that the Atlantic Meridional Overturning Circulation (AMOC) changes in OrbIce are similar to those in All_forcings (see Fig.1 in the supplement file Figure_rev1.pdf). This systematic response, observed across a large number of simulations, suggests that the inclusion of ice-sheet variations in the set of forcings induces a consistent response in the ocean circulation. This justifies the exploration of the relationship between AMOC variations and temperatures, despite the short run time of the snapshots. As such, Fig.2 and Fig.3 in the supplement file Figure_rev1.pdf show the relationship between AMOC strength and hemispheric sea surface temperatures (SHΔSST and NHΔSST) for the 4 simulation sets All_forcings, OrbGhg, OrbIce and OrbOnly. Pearson’s correlation tests results are indicated for the three time slices preMPT, MPT and postMPT.
  These plots show a strong negative correlation between AMOC strength and SHΔSST in both All_forcings and OrbIce, with r=-0.67 after the MPT for both snapshot sets (see Fig.2 in the supplement file Figure_rev1.pdf). The strength of this relationship increases relative to the pre-MPT and MPT intervals, with r=-0.22 and r=-0.28 for All_forcings and OrbIce, respectively, during the pre-MPT, and r=-0.26 and r=-0.34, respectively, during the MPT. In contrast, AMOC strength and SHΔSST are positively correlated in OrbGhg, with r=0.57 pre-MPT, r=0.41 during the MPT, and r=0.57 after the MPT.
  OrbOnly also shows an increase in the strength of the linear correlation through time, although both variables remain more weakly correlated than in the other simulation sets.
  Interestingly, the relationship between temperature and AMOC is weaker for NHΔSST when ice-sheet variations are included in the set of forcings. In All_forcings, for example, r=0.38 pre-MPT and r=-0.35 after the MPT, while the correlation is not significant during the MPT. Notably, after the MPT, the correlation between AMOC strength and NHΔSST is also not significant in OrbIce (see Fig. 3 in the supplement file Figure_rev1.pdf). Conversely, in OrbGhg and OrbOnly, the relationship between AMOC strength and SHΔSST is stronger than that with NHΔSST (see Fig. 2 & Fig.3 n the supplement file Figure_rev1.pdf).
  These results highlight the strong relationship between AMOC strength and SHΔSST induced by the inclusion of ice-sheet variations in the forcings, which supplants the opposite relationship induced by greenhouse gases variations. Further exploration on the mechanisms relating these two variables will be the subject of future work. As it is not part of the main results show in the current paper, we have removed from the abstract any conclusions about it. If useful, we can include these analysis to the supplementary of the paper.
  The effects of greenhouse gases and ice sheets on temperature may depend on the background climate state. The manuscript would therefore benefit from comparisons among several representative climate states, including a warm Pliocene state and pre- and post-MPT glacial and interglacial states. For each state, the authors should examine not only the magnitude of the temperature response but also the underlying physical processes. Such analyses would help determine whether similar changes in greenhouse gases or ice sheets produce different temperature responses under different background conditions and would identify the mechanisms responsible for any state-dependent climate sensitivity.
  
  Thank you for highlighting the influence of the background climate on its sensitivity to changes in ice sheets and greenhouse gases. This is precisely what we intended to explore by examining the non-linear response to the additive forcings through time with Figure 7 in the initial manuscript. However, as pointed out by Reviewer 2, this was not clearly demonstrated. We have therefore modified our approach by assessing the extent to which the response of OrbGhg + OrbIce − OrbOnly diverges from that of All_forcings in terms of both trends and glacial–interglacial amplitudes, before, during and after the MPT. The complete analysis can be found in the response to Reviewer’s 2 comments.
  We understand that this kind of analysis only focuses on the temperature response to ice sheet or greenhouse gases changes and that we are not exploring in depth the physical pathways behind each of the 919 climate states we have. The title has been reformulated to better convey the nature of the results we provide, and to avoid any misunderstanding about explaining which physical mechanisms triggered the MPT. As stated on lines 282-287, our work differs from other modelling studies which investigate singular climate states (which have been done multiple times with HadCM3 already), such as the Late Glacial Maximum (see Abe-Ouchi et al., 2015 for impact of different ice sheets on the radiative balance) or the Pliocene (see Hunter et al. 2019 for HadCM3 in particular). This type of study generally investigates the response of different ice sheet or greenhouse gases forcings to a same climate state, potentially in comparison with the pre-industrial to assess changes in climate sensitivity. Here, we are focusing of the evolving response of temperature and we do not think that looking at five individual climate states would really help to uncover the physical pathways responsible for the temperature evolution over the 3.6 Ma. To do so, we would need to investigate these potential pathways over a same time-scale, as we started to do with the relationship between the AMOC and the Southern temperatures.
  References:
  Berends, C. J., Köhler, P., Lourens, L. J., & van de Wal, R. S. W. (2021). On the cause of the mid-Pleistocene transition. Reviews of Geophysics, 59, e2020RG000727. https://doi.org/10.1029/2020RG000727
  Abe-Ouchi, A. and Saito, F. and Kageyama, M. and Braconnot, P. and Harrison, S. P. and Lambeck, K. and Otto-Bliesner, B. L. and Peltier, W. R. and Tarasov, L. and Peterschmitt, J.-Y. and Takahashi, K. (2015). Ice-sheet configuration in the CMIP5/PMIP3 Last Glacial Maximum experiments. Geoscientific Model Development. 10.5194/gmd-8-3621-2015
  Hunter, S. J. and Haywood, A. M. and Dolan, A. M. and Tindall, J. C.(2019). The HadCM3 contribution to PlioMIP phase 2. Climate of the Past. https://cp.copernicus.org/articles/15/1691/2019/
  Joy S. Singarayer, Paul J. Valdes, High-latitude climate sensitivity to ice-sheet forcing over the last 120kyr, Quaternary Science Reviews, https://doi.org/10.1016/j.quascirev.2009.10.011.
  Armstrong, E., Izumi, K. & Valdes, P. Identifying the mechanisms of DO-scale oscillations in a GCM: a salt oscillator triggered by the Laurentide ice sheet. Clim Dyn 60, 3983–4001 (2023). https://doi.org/10.1007/s00382-022-06564-y
  
  Citation: https://doi.org/10.5194/egusphere-2026-3719-AC1
RC2:
'Comment on egusphere-2026-3719', Anonymous Referee #2, 06 Aug 2026
Review of 'Relative contribution of CO2 and ice sheets on the Mid-Pleistocene Transition' by Jeanne Millot-Weil et al.
In this manuscript the authors present a comprehensive new set of snapshot climate model simulations covering the last 3.6 million years, sampled at 4 kyr resolution. They run 4 sets of simulations with different combinations of time-dependent external forcings (orbital parameters, GHGs and continental ice sheet) that allow to separate the role of different forcings on the simulated climate trends and variability. The analysis is focused on global and hemispheric temperature changes and the results show that, when all forcings are included, the model generally reproduces long-term global temperature trends and glacial-interglacial variability reasonably well. The results provide new insights into the impact of different forcings in shaping the temperature response across the Mid-Pleistocene Transition. The paper is generally well written and the methods and results clearly presented. The labelling and the quality of the figures need to be improved.

I support publication of the paper in Climate of the Past after the comments below have been addressed.

Main comments
The title and the abstract of the paper give the impression that the paper explores the role of CO2 and ice sheets in driving the change in periodicity of the global climate from ~40 to ~100 kyrs across the Mid-Pleistocene transition, which is not supported by the main paper content, where the authors clearly state that conclusions about the origin of the MPT can not be drawn from the present set of simulations. Please rephrase the abstract to make this clear. In particular, the sentence ending with ‘…show that greenhouse gases exert a direct and dominant role on both the long-term global cooling trend and the shift in glacial-interglacial cyclicity associated with the MPT’ can be interpreted as stating that GHGs (CO2) caused the MPT. Also, I suggest reformulating the title to something like ‘Relative contribution of CO2 and ice sheets to temperature trends and variability across the Mid-Pleistocene Transition’.
The abstract could instead better summarise the main findings:
GHGs dominate long-term temperature trends

Ice sheets and GHGs have a comparable impact on glacial-interglacial variations, with ice sheets being more important for the pre-MPT and GHGs for the post-MPT

The prescribed changes in the periodicity of the GHGs and ice sheets forcings across the MPT contribute roughly equally to the transition from 40 to 100 kyrs cyclicity in global temperature

Something on the difference in NH vs SH response in ST and SST at the 40 and 100 ka scales

The last sentence in the abstract is about the amplitude of the glacial-interglacial cycles and seems to suggest that the appearance/disappearance of ice sheets (mainly over the NH land) have no direct effect on (global) temperature, only indirectly through sea ice and ocean circulation. I would expect at least the ice sheet surface albedo to have an impact. Moreover, sea level changes due to changes in land ice volume could also impact the carbon cycle and atmospheric CO2.
How non-linearly the different forcings combine to generate the All_forcings response is an interesting aspect of the analysis. However, the analysis is not cleanly done, as it accounts multiple times for orbital forcing. This is particularly relevant for the case of the glacial-interglacial amplitudes (e.g. Fig. 7b). A more appropriate decomposition to test non-linearities could be: All_forcings ~=(?) OrbGhg + OrbIce - OrbOnly. On a related note, why does Orb show up only in panel c of Fig. 7a?

For the trend analysis in Fig. 7a, could you elaborate a bit more on why the post-MPT is qualitatively different?

Minor comments
L. 51-53: This sentence is not very clear. You have 3 different time-dependent forcings that are combined in 4 different sets of simulations.
L. 59: ‘stability of the ice sheet’ -> ‘ice sheet dynamics’?
L. 91: Fig.1 (a) A) etc. is not standard figure panel indexing.
L. 104: ‘…forcings, here,…’ -> ‘…forcings. Here,…’
L. 105: ‘parts’ -> ‘sections’?
L. 110-111: Not everyone might be familiar with the concept of CO2 equivalent. I would suggest to add one sentence to explain what it is.
L. 111: ‘first’ or ‘last’?
L. 125: ‘all orbital scale parameters’ is not very clear to me. Maybe ‘all time-dependent forcings’ instead?
In Table 1, specify that ‘Constant’ means ‘Constant pre-industrial’
L. 140: Not just the orbital forcing, but also the other forcings were accelerated.
In Fig. 1 (b) A) the unit is wrong and the global ice volume is not the sum of the NH and SH components
L. 169: ‘before and after’ -> ‘before’
L. 170: L. 78: °C/ka is indicated as the unit for the temperature trends, but that can’t be right.
L. 288: Why ‘comparable’?
L. 306-307: I don’t fully understand this sentence.
L. 332: ‘…ice sheet variations show greater amplitude of variations than the greenhouse gases’; is this true? And how does this affect the long-term trends discussed here?
L. 333-342: I have to admit that I couldn’t really follow the reasoning here.
Fig. 8 shows normalized power spectra, which give the impression that OrbOnly shows the largest response to e.g. precession. It would possibly be more useful to see the un-normalized spectra, to make them directly comparable.
L. 406: ‘…the obliquity pacing diminishes when the 100-ka period of the MPT emerges’; where can this be seen?
L. 523: ‘sense’?
The sign convention on the trends in Fig. A2 is opposite to the one used in the main text.
Citation: https://doi.org/10.5194/egusphere-2026-3719-RC2
- AC2:
  'Reply on RC2', Jeanne Millot-Weil, 29 Sep 2026
  In this manuscript the authors present a comprehensive new set of snapshot climate model simulations covering the last 3.6 million years, sampled at 4 kyr resolution. They run 4 sets of simulations with different combinations of time-dependent external forcings (orbital parameters, GHGs and continental ice sheet) that allow to separate the role of different forcings on the simulated climate trends and variability. The analysis is focused on global and hemispheric temperature changes and the results show that, when all forcings are included, the model generally reproduces long-term global temperature trends and glacial-interglacial variability reasonably well. The results provide new insights into the impact of different forcings in shaping the temperature response across the Mid-Pleistocene Transition. The paper is generally well written and the methods and results clearly presented. The labelling and the quality of the figures need to be improved.
  
  I support publication of the paper in Climate of the Past after the comments below have been addressed.
  Thank you very much for your insights and helpful suggestions, we have made changes to the manuscript accordingly.
  Main comments
  The title and the abstract of the paper give the impression that the paper explores the role of CO2 and ice sheets in driving the change in periodicity of the global climate from ~40 to ~100 kyrs across the Mid-Pleistocene transition, which is not supported by the main paper content, where the authors clearly state that conclusions about the origin of the MPT can not be drawn from the present set of simulations. Please rephrase the abstract to make this clear. In particular, the sentence ending with ‘…show that greenhouse gases exert a direct and dominant role on both the long-term global cooling trend and the shift in glacial-interglacial cyclicity associated with the MPT’ can be interpreted as stating that GHGs (CO2) caused the MPT. Also, I suggest reformulating the title to something like ‘Relative contribution of CO2 and ice sheets to temperature trends and variability across the Mid-Pleistocene Transition’.
  The abstract could instead better summarise the main findings:
  GHGs dominate long-term temperature trends
  
  Ice sheets and GHGs have a comparable impact on glacial-interglacial variations, with ice sheets being more important for the pre-MPT and GHGs for the post-MPT
  
  The prescribed changes in the periodicity of the GHGs and ice sheets forcings across the MPT contribute roughly equally to the transition from 40 to 100 kyrs cyclicity in global temperature
  
  Something on the difference in NH vs SH response in ST and SST at the 40 and 100 ka scales
  
  The last sentence in the abstract is about the amplitude of the glacial-interglacial cycles and seems to suggest that the appearance/disappearance of ice sheets (mainly over the NH land) have no direct effect on (global) temperature, only indirectly through sea ice and ocean circulation. I would expect at least the ice sheet surface albedo to have an impact. Moreover, sea level changes due to changes in land ice volume could also impact the carbon cycle and atmospheric CO2.
  Thank you for these remarks.
  We have modified the title to better conveys the nature of the results presented in the manuscript: “Relative contributions of CO2 and ice sheets to temperature trends and variability over the past 3.6 Ma, with insights on the Mid-Pleistocene Transition”.
  In addition, we agree that the abstract was indeed misleading, and needed to be revised to better summarize the main findings and avoid inaccurate interpretation about the physical pathways through which ice sheet impacts the temperatures. The new version we propose following your suggestions is as follows (the edits are indicated by bold characters):
  
  "The Mid-Pleistocene Transition (MPT) represents one of the most prominent and debated climate reorganizations within the long-term cooling trend of the last 3.6 million years. Although the role of ice sheets, greenhouse gases and their combined effects have all been implicated, their individual contributions to the MPT remain unresolved.
  Here, we assess their respective roles using a unique set of four series of fully coupled equilibrium palaeoclimate model simulations spanning the last 3.6 Ma at 4-ka intervals, constrained by realistic boundary conditions. These simulations differ in their prescribed insolation, greenhouse gases and ice sheet forcings where a realistic baseline time evolving scenario is analysed against a series where each forcing is kept static throughout to constrain their individual impact on the MPT.
  First, we evaluate the performance of the four sets of simulations including the combination of the three changing forcings against geological data. We find that the model captures the main characteristics of temperature variations identified by geological records, such as the global cooling trend over the past 3.6 Ma, acceleration of cooling at the MPT, with amplification of the glacial-interglacial cycles along with a change of pace from a 40-ka to a 100-ka cyclicity. Then, we compare the four simulated timeseries to disuntangle the individual role of each forcing through time. We show that greenhouse gases exert a dominant role on long-term temperature trends, whereas both contributions of ice sheet and greenhouse gases variability have comparable impact on the amplitude of glacial-interglacial variability. However, ice-sheet variations exert a stronger influence on glacial-interglacial amplitude prior to the MPT, whereas greenhouse gases emerge as the main driver after the MPT. In addition, the changes in the periodicity of greenhouse gases and ice sheet both contribute to the transition from 40-ka to 100-ka glacial-interglacial cyclicity at the global scale and for both air and sea surface temperatures, but regional responses reveal important differences. Greenhouse gases consistently impact air and sea surface temperature variability across both hemispheres, whereas ice-sheet forcing induces a weaker 40-ka periodicity in sea surface temperatures than in air surface temperatures in/across both hemispheres, and the 100-ka periodicity shift is absent from Northern sea surface temperature variability."
  How non-linearly the different forcings combine to generate the All_forcings response is an interesting aspect of the analysis. However, the analysis is not cleanly done, as it accounts multiple times for orbital forcing. This is particularly relevant for the case of the glacial-interglacial amplitudes (e.g. Fig. 7b). A more appropriate decomposition to test non-linearities could be: All_forcings ~=(?) OrbGhg + OrbIce - OrbOnly. On a related note, why does Orb show up only in panel c of Fig. 7a?
  Orb shows up only in panel c of the initial version of Figure 7a because only during the MPT the simulation set Orb shows temperature trends that differ from 0 (as shown on Table A2).
  For the trend analysis in Fig. 7a, could you elaborate a bit more on why the post-MPT is qualitatively different?
  Thank you for your suggestions on how to better handle the exploration of non-linearities between forcings. We have modified Fig.7 a & b, and Tab.A1 and A2 to show the deviation of the combination OrbGhg + OrbIce – Orb to the results obtained with All_forcings (see Figure 1, Tab. A1 and Tab.A2 in the supplementary file Figure_rev2.pdf).
  For the temperature trends, Figure 1(a) in the supplementary file Figure_rev2.pdf shows results consistent with those presented in Figure 7a of the initial manuscript. Before the MPT, the linear additivity of the time-varying forcings appears to hold, with OrbGhg + OrbIce − Orb showing only minor deviations from All_forcings (a maximum deviation of 5% for the Northern Hemisphere sea surface temperature trend). However, this hypothesis is no longer verified during the MPT, when OrbGhg + OrbIce − Orb globally overestimates the All_forcings trends (overestimation of 11% and 25% for global mean surface and sea surface temperature trends respectively), nor after the MPT, when it substantially underestimates them (underestimation of ~13% at global scale).
  The overestimation of the trends at the MPT mainly originates from Northern hemisphere trends, where the deviations are the highest (more than 20%, see Tab.A2), whereas Southern hemisphere trends seem to still respond linearly to the combination of forcings as prior to the MPT (less than 3%, see Tab.A2). Deviations of Tropical temperature trends sit in between both hemispheres' results (~12%, see Tab.A2). Since the greenhouse gas trend does not change dramatically between the preMPT and MPT time slices (see Fig.1(a) from the initial manuscript), this result suggests that an important role is being played by the growing Northern ice sheet during the MPT (see Fig.1(b) from the initial manuscript). In particular, it indicates a different greenhouse gas sensitivity of the Northern hemisphere between the preMPT climate state and the ongoing MPT climate state.
  After the MPT, the deviations of the combination OrbGhg + OrbIce – Orb from All_forcings appear particularly important (for example, an underestimation of 35% for Tropical SST). These numbers should be interpreted with caution, as the temperature trends after the MPT are substantially weaker than for the preceding periods (and the Tropics are the region of the weakest trends), and thus even small absolute differences can result in relatively large percentage deviations. Incidentally, over the full 3.6 Ma, OrbGhg + OrbIce – Orb overestimates the trends relative to All_forcings. In addition, after the MPT, greenhouse gases do not decrease steadily anymore but oscillate around constant value (~250 ppm). The contribution of OrbGhg in explaining the trends over this time period is therefore reduced compared to the previous time slices (see Tab.A2 in the supplementary file Figure_rev2.pdf). As a result, a quantitative interpretation of the deviations of postMPT OrbGhg + OrbIce – Orb trends from All_forcings trends, as a comparison with the previous time slices, might be misleading. Nevertheless, the qualitative shift from an overestimation to an underestimation of the trends by the linear combination of forcings is still noteworthy, as it suggests that interactions between the different forcings may vary across climate states, suggesting a change of regime at the MPT.
  The mean absolute deviations from the mean were also calculated for the time series from the OrbGhg + OrbIce − Orb combination and compared with those obtained for All_forcings to assess whether the glacial–interglacial amplitudes are reproduced by linear additivity of the forcings (see Figure 1 (b) in the supplementary file Figure_rev2.pdf). As for the trends, the three time slices show distinct results, but OrbGhg + OrbIce − Orb consistently tends to overestimate the amplitude of the glacial-interglacial variability.
  More precisely, before the MPT, when glacial–interglacial amplitudes were smaller than during more recent periods, OrbGhg + OrbIce − Orb only slightly overestimates the amplitude of variability in All_forcings, by less than 10% (see Tab A1 in the supplementary file Figure_rev2.pdf). During the MPT, linear additivity reproduces the amplitude of variability in tropical and Southern hemisphere temperatures well, with differences of less than 2% (see Tab A1 in the supplementary file Figure_rev2.pdf). Conversely, OrbGhg + OrbIce − Orb overestimates the glacial–interglacial amplitude in Northern hemisphere temperatures, by 10% and 14% for air and sea surface temperatures respectively (see Tab A1 in the supplementary file Figure_rev2.pdf). The only exception is global mean SST, for which OrbGhg + OrbIce − Orb underestimates the amplitude by 5%, but the mean absolute deviations from the mean are very similar in the two cases (0.39°C and 0.41°C for OrbGhg + OrbIce − Orb and All_forcings, respectively, see Tab A1 in the supplementary file Figure_rev2.pdf), suggesting that the linear additivity seems to hold at the global scale. After the MPT, the overestimation of Northern hemisphere variability becomes more pronounced, with OrbGhg + OrbIce − Orb exceeding the All_forcings amplitude by 14% and 26% for air and sea surface temperatures respectively (see Tab A1 in the supplementary file Figure_rev2.pdf), whereas the overestimation decreases towards southern latitudes (less than 9% for the Tropics and less than 4% for the Southern hemisphere). These deviations emerge as Northern hemisphere ice sheet grows, suggesting that it exerts an important influence on the temperature response to the combined forcings. The results appear particularly pronounced for Northern sea surface temperatures compared with surface air temperatures. However, the glacial–interglacial amplitude is smaller for SST than for surface air temperature (see Tab. A1 in the supplementary file Figure_rev2.pdf), which may again artificially amplify the relative deviations.
  Overall, the comparison between OrbGhg + OrbIce − Orb and All_forcings, both in trends and glacial-interglacial amplitude, suggests that the growth of Northern hemisphere ice sheets plays an important role in modulating temperature sensitivity to individual forcings and their combined effects.
  Minor comments
  51-53: This sentence is not very clear. You have 3 different time-dependent forcings that are combined in 4 different sets of simulations.
  The sentence has been modified as follows to improve clarity: “Four unique sets of 919 simulations (spaced approximately every 4000 years) have been performed with different combinations of prescribed insolation, realistic ice sheet and greenhouse gas concentration evolutions.”
  59: ‘stability of the ice sheet’ -> ‘ice sheet dynamics’?
  “Stability of the ice sheet” has been modified with “ice sheet dynamics”.
  91: Fig.1 (a) A) etc. is not standard figure panel indexing.
  The indexing has been changed from (A), (B), etc. to (i), (ii), etc. to follow a more standard convention.
  104: ‘…forcings, here,…’ -> ‘…forcings. Here,…’ Ok
  105: ‘parts’ -> ‘sections’? Ok
  110-111: Not everyone might be familiar with the concept of CO2 equivalent. I would suggest to add one sentence to explain what it is.
  The following sentence has been added to define what CO2 equivalent is: “CO2_eq represents the warming effect of all greenhouse gases, expressed in units of CO2.”
  111: ‘first’ or ‘last’?
  It was pointing the last 800 ka indeed. Thank you for spotting this.
  125: ‘all orbital scale parameters’ is not very clear to me. Maybe ‘all time-dependent forcings’ instead?
  “all orbital scale parameters” has been modified to “all time-dependent forcings”
  In Table 1, specify that ‘Constant’ means ‘Constant pre-industrial’ Changed.
  140: Not just the orbital forcing, but also the other forcings were accelerated.
  Sentence modified to “Their transient approach focused on accelerating the orbital year by a factor of five, thereby accelerating all time-dependent parameters (ice sheet, greenhouse gases and insolation).”
  In Fig. 1 (b) A) the unit is wrong and the global ice volume is not the sum of the NH and SH components
  Thank you for spotting this. Figure 2 in this document shows the updated Figure. 2 in the supplementary file Figure_rev2.pdf
  169: ‘before and after’ -> ‘before’ Fixed.
  170: L. 78: °C/ka is indicated as the unit for the temperature trends, but that can’t be right.
  Yes, it is °C/Ma, thank you for spotting this.
  288: Why ‘comparable’? This has been removed.
  306-307: I don’t fully understand this sentence. This sentence indeed does not reflect the point we wanted to make. Here is the reformulation we propose:
  “This difference is strongly influenced by the Pliocene temperatures, for which OrbGhg closely reproduces the All_forcings results, whereas OrbIce does not.”
  332: ‘…ice sheet variations show greater amplitude of variations than the greenhouse gases’; is this true? And how does this affect the long-term trends discussed here?
  Thank you for spotting this. Comparing amplitude of variations between greenhouse gases and ice volume does not provide anything useful to characterise the long-term trends discussed in this section. Here is the revised version of the sentence:
  “…during the pre-MPT interval, greenhouse gas concentrations decrease more rapidly than ice sheets grow, whereas after the MPT, greenhouse gas concentrations stabilise while ice sheets continue to expand.”
  333-342: I have to admit that I couldn’t really follow the reasoning here.
  These lines were indeed difficult to follow. We have therefore rewritten them to improve clarity and avoid overinterpretation:
  “Interestingly, during the MPT, the trend slopes in both OrbGhg and OrbIce are generally closer to those observed in All_forcings than during pre or post-MPT intervals. The growth of the ice sheets during this period (Fig.1(a)(ii) and (b)(i)) can explain why OrbIce more closely reproduces the All_forcings trends. However, the absence of a substantial change in the CO2eq trend relative to the pre-MPT interval (Fig.1 (a)(iii)) does not readily explain why OrbGhg also provides a better match to the All_forcings trend. Instead, this could suggest that the pre-industrial ice-sheet configuration prescribed in OrbGhg is more representative of the climate state during the MPT than before the MPT. “
  Fig. 8 shows normalized power spectra, which give the impression that OrbOnly shows the largest response to e.g. precession. It would possibly be more useful to see the un-normalized spectra, to make them directly comparable.
  We reproduced Fig.8 without normalizing the timeseries (see Figure 3 in this document). The results do not fundamentally change but OrbOnly shows indeed less strong precession peaks compared to the other timeseries.
  406: ‘…the obliquity pacing diminishes when the 100-ka period of the MPT emerges’; where can this be seen? This can be seen on Fig. 1 (a) D) of the initial manuscript and renamed Fig.1 (a) iv).
  523: ‘sense’? This has been removed.
  The sign convention on the trends in Fig. A2 is opposite to the one used in the main text. Fixed.
  
  Citation: https://doi.org/10.5194/egusphere-2026-3719-AC2

Jeanne Millot-Weil, Paul Valdes, and Alexander Farnsworth

Viewed

Total article views: 395 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
241	120	34	395	18	16

HTML: 241
PDF: 120
XML: 34
Total: 395
BibTeX: 18
EndNote: 16

Views and downloads (calculated since 08 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	120	51	11	182
Aug 2026	38	18	4	60
Sep 2026	66	42	15	123
Oct 2026	17	9	4	30

Cumulative views and downloads (calculated since 08 Jul 2026)

Month	HTML	PDF	XML	Total
Jul 2026	120	51	11	182
Aug 2026	38	18	4	60
Sep 2026	66	42	15	123
Oct 2026	17	9	4	30

Viewed (geographical distribution)

Total article views: 306 (including HTML, PDF, and XML) Thereof 306 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 07 Oct 2026

Short summary

The Mid-Pleistocene Transition (MPT, ~1 million years ago) marked a major shift in glacial–interglacial cycles, linked to interactions between ice sheet and greenhouse gases (Ghg). Their individual contributions to glacial-interglacial changes during this period is unclear. Using climate model experiments, we untangle their respective impacts and show that Ghg impact directly surface temperature at the MPT, whereas ice sheets mainly influence ocean circulation changes.


Total:	0
HTML:	0
PDF:	0
XML:	0

Relative contributions of CO2 and ice sheets on the Mid-Pleistocene Transition

Viewed

Viewed (geographical distribution)

Relative contributions of CO₂ and ice sheets on the Mid-Pleistocene Transition