| 18 Aug 2026
Enhanced ocean mixing improves proxy-model agreement for an Early Pliocene seasonally ice-free Arctic Ocean
Abstract. The Arctic Ocean rapidly transitions towards ice-free summers, a "blue" (i.e., summer sea ice-free) state driven by Arctic amplification of abrupt global warming. To improve climate projections and better understand the consequences of reduced Arctic sea ice in a warming climate, it is essential to examine past warm intervals such as the Early Pliocene (~5–4 Ma), when diminished Arctic sea ice coincided with higher global temperatures. Proxy records from the Arctic-Atlantic Gateway (AAG) (~70–80° N), combined with climate model simulations, indicate periodically ice-free summers during this time. Yet, model-data comparisons for the central Arctic Ocean (>80° N) reveal winter sea ice as a persistent feature. Here, we show that proxy evidence for a "blue" Arctic is best reproduced when enhanced vertical ocean mixing is integrated into climate models. Under enhanced mixing, Arctic warming occurs through different mechanisms depending on the model geography, whereas simulations with standard mixing show substantially weaker agreement with proxy evidence. We do not interpret enhanced mixing as the unique solution but rather as a physically motivated sensitivity test that represents unresolved ocean processes. Our results demonstrate that modifying a single ocean parameterization within physically plausible boundaries substantially improves proxy-model agreement for a warmer than modern Early Pliocene, both at the AAG and for the wider Arctic. Combined with published proxy evidence for a warmer Arctic Ocean interior, our findings underscore the need for improved representation of ocean dynamics in paleoclimate simulations. Next-generation climate models should evaluate the sensitivity of simulated "blue" Arctic climates to ocean mixing alongside other key processes, including cloud microphysics and radiative feedbacks, to increase confidence in projections of seasonally ice-free Arctic conditions
General Comments
Arctic data-model mismatch is a pervasive problem that persists across multiple warm-period model intercomparison projects and experimental phases. This manuscript addresses that important discrepancy by asking whether enhanced ocean mixing can help reconcile the model outputs against proxy evidence for Arctic warmth during the Early Pliocene. The study combines previously published proxy records with climate-model experiments bracketing the Early Pliocene and presents a framework linking ocean mixing, heat transport, sea-ice extent, and Arctic amplification. The topic is of broad relevance to both the palaeoclimate and climate-modelling communities.
The manuscript raises an interesting and worthwhile hypothesis regarding the role of unresolved ocean processes in past warm climates and demonstrates that, what seems to be a relatively modest perturbation to ocean-mixing parameterization can substantially improve agreement between simulations and a select number of Arctic terrestrial and marine proxy records. The discussion of contrasting heat-transport pathways in the Miocene and Pliocene model configurations is also a valuable contribution.
My main concerns are: the rationale for an Early Pliocene DMC, incomplete use of Early Pliocene terrestrial data for DMC, and a broad need for expression revision throughout to improve clarity.
Overall, I consider this a valuable contribution that merits publication after the authors address the specific comments below.
Specific Comments
The rationale for why an Early Pliocene DMC is conducted, as opposed to a Miocene and a Late Pliocene study.
Comparison to Early Pliocene proxy evidence for a seasonally ice-free Arctic Ocean appears to be the primary motivation for both the bracketing approach and the focus on the Early Pliocene. If so, this should be stated explicitly from the outset. The manuscript explains why Miocene and mPWP simulations are used to bracket Early Pliocene conditions, but it is less clear why the Early Pliocene is selected as the primary target interval rather than conducting more conventional model-proxy comparisons within the simulated intervals themselves. If the rationale is that geological evidence for a seasonally ice-free ("blue") Arctic is strongest during the Early Pliocene, and that this interval therefore provides the most informative test of model performance and proxy-model agreement, this should be stated explicitly early in the manuscript.
Additionally, the manuscript generally frames the Miocene and mid-Piacenzian experiments appropriately as end-member simulations used to explore Early Pliocene conditions. However, in some places the distinction between the simulated intervals and the target Early Pliocene interval becomes blurred.
Terrestrial model-data comparison appears incomplete
The manuscript's model-data comparison for the wider Arctic relies on only two terrestrial sites: Beaver Pond and Lake El'gygytgyn (lines 89-90; lines 680-698; Fig. 8) as part of the authors claim that enhanced ocean mixing improves agreement "across the wider Arctic realm"
The Lake El'gygytgyn record only beings at the start of the Late Pliocene and does not compare directly to the Early Pliocene. It is currently labelled EP.
Presuming the authors only wish to include sites above >66°N, Tnekveem, an EP site in Siberia, would be useful. If considering sites down to 60ºN, a few other in the same publication, below, could be included.
Late Miocene sites for comparison include: Yanran, Temmirdekh-khaj, Omoloy, and Nekkeveem /early Chattian.
Popova, S., Utescher, T., Gromyko, D., Bruch, A., Mosbrugger, V., 2012. Palaeoclimate Evolution in Siberia and the Russian Far East from the Oligocene to Pliocene – Evidence from Fruit and Seed Floras. TURKISH JOURNAL OF EARTH SCIENCES. 10.3906/yer-1005-6
This list is not likely comprehensive.
I was also surprised not to see other Canadian High Arctic terrestrial records that have been interpreted as Early Pliocene deposits with reconstructions of Arctic warmth, including Banks Island-Beaufort Formation, Prince Partick Island, Fyles Leaf Beds, and more recent papers on Beaver Pond. The absence of Meighen Island, was particularly notable given its relevance to the sea ice discussion:
“In colder environments, Arctica occupies shallower water, so its presence on Meighen Island probably indicates that shelf temperatures at 25–45 m and probably sea- surface temperatures did not fall below 0 °C at any time of the year. This is evidence that the coastal areas of the landmass now partly represented by the Beaufort Formation on Meighen Island did not possess perennial ice cover but more likely had a short season of sea ice.” Barendregt et al. 2021.
Additional Late Pliocene Arctic sites include Niguanak and Bluefish.
If they have not yet, the authors ought to consider the following terrestrial evidence:
Barendregt, R.W., Matthews, J.V., Jr, Behan-Pelletier, V., Brigham-Grette, J., Fyles, J.G., Ovenden, L.E., McNeil, D.H., Brouwers, E., Marincovich, L., Fletcher, T., 2021. Biostratigraphy, age, and the paleoenvironment of the Pliocene Beaufort Formation on Meighen Island, Canadian Arctic Archipelago. Geol. Soc. Am. Special Paper 551, 1–39. https://doi.org/10.1130/2021.2551(01)
Carter, L.D., Christensen, D., Hopkins, D.M., Matthews Jr, J.V., Nelson, R.E., Ovenden, L.E., Rybczynski, N., Fletcher, T.L., 2024. Stratigraphy and paleontology (plant and arthropod fossils) from the Late Neogene Niguanak site, Arctic Slope, Northern Alaska. Arct. Antarct. Alp. Res. 56, 2407714. https://doi.org/10.1080/15230430.2024.2407714
Elias, S.A., Matthews, J.V., Jr., 2002. Arctic North American seasonal temperatures from the latest Miocene to the Early Pleistocene, based on mutual climatic range analysis of fossil beetle assemblages. Can. J. Earth Sci. 39, 911–920. https://doi.org/10.1139/e01-096
Fletcher, T., Feng, R., Telka, A.M., Matthews, J.V., Jr., Ballantyne, A., 2017. Floral dissimilarity and the influence of climate in the Pliocene High Arctic: Biotic and abiotic influences on five sites on the Canadian Arctic Archipelago. Front. Ecol. Evol. 5, 19. https://doi.org/10.3389/fevo.2017.00019
Fletcher, T.L., Csank, A.Z., Ballantyne, A.P., 2019. Identifying bias in cold season temperature reconstructions by beetle mutual climatic range methods in the Pliocene Canadian High Arctic. Palaeogeogr Palaeoclimatol Palaeoecol 514, 672–676. https://doi.org/10.1016/j.palaeo.2018.11.025
Fletcher, T.L., Warden, L., Sinninghe Damsté, J.S., Brown, K.J., Rybczynski, N., Gosse, J.C., Ballantyne, A.P., 2019. Evidence for fire in the Pliocene Arctic in response to amplified temperature. Clim. Past 15, 1063–1081. https://doi.org/10.5194/cp-15-1063-2019
Fletcher, T., Eble, C., Sinninghe Damsté, J.S., Brown, K.J., Rybczynski, N., Gosse, J., Liu, Z., Ballantyne, A., 2021. Widespread wildfire across the Pliocene Canadian Arctic Archipelago. Palaeogeogr Palaeoclimatol Palaeoecol 584, 110653. https://doi.org/10.1016/j.palaeo.2021.110653
Porter, T.J., Anhäuser, T., Halfar, J., Keppler, F., Csank, A.Z., Williams, C.J., 2022. Canadian Arctic Neogene Temperatures Reconstructed From Hydrogen Isotopes of Lignin-Methoxy Groups From Sub-Fossil Wood. Paleoceanogr Paleoclimatol 37, e2021PA004345. https://doi.org/10.1029/2021PA004345
Tindall, J., Haywood, A., Salzmann, U., Dolan, A., Fletcher, T., 2022. The Warm Winter Paradox in the Pliocene Northern High Latitudes. Clim. Past 18, 1385–1405. https://doi.org/10.5194/cp-18-1385-2022
Although this research focuses on sea ice, limited Early Pliocene evidence mean the terrestrial records form an important component of the broader Arctic comparison, thus the authors should clarify the criteria used to select sites for inclusion in Figures 1 and 8 and discuss whether inclusion of additional Arctic terrestrial records would affect their conclusions.
Specifics by line
Line 94: I do not think the cited literature supports the statement that the Pliocene was characterised by "pronounced glacial-interglacial cycles". McClymont et al. (2023), the only citation directly addressing this issue, instead emphasise the progressive development of larger-amplitude climate as particularly following the mPWP, with events from around the start of the Late Pliocene, such as M2 representing exceptions rather than the norm. The current wording gives the impression of Pleistocene-style and scale glacial-interglacial cycling, which does not accurately reflect the current understanding of the generally lower-amplitude climate oscillations during much of the Pliocene, and especially the Early Pliocene. A description reflecting the generally lower-amplitude variability than the late Pliocene and Pleistocene would be aligned with the current literature.
(See also Line 183-186 here, and the IRD record in Figure 2, which appears to conflict).
Line 125: This statement appears to imply that the Early Pliocene differed from the Late Pliocene primarily through higher atmospheric CO₂ concentrations and relatively limited palaeogeographic changes. However, the PlioMIP3 Early Pliocene experiment was designed as a sensitivity experiment incorporating two key differentiating features (increased CO₂ and an open Central American Seaway) while intentionally retaining Late Pliocene boundary conditions elsewhere. It was not intended to represent all likely differences between the Early and Late Pliocene. Did the authors mean that existing Early Pliocene modelling experiments differ from Late Pliocene simulations principally in these respects, rather than implying that these were the only important climatic and palaeogeographic differences between the two intervals?
Line 130: alongside other key boundary condition changes between the mPWP and Miocene Climatic Optimum?
Line 145: Do we have evidence or past studies that suggest vertical mixing parameters based on modern observation don’t or may not hold?
Line 146: I think this is very unclear to a reader, who hasn’t read this Method yet. How are two models an analogue for one.
Line 146–162: This is results and discussion. I don’t think it belongs in the introduction.
Line 167: Needs support.
Line 175: Salzmann et al 2011 is a review focused on the Late Pliocene. It is not a synthesis of Early Pliocene global palaeoenvironments. "The review focuses on the Piacenzian (3.6-2.58 Ma) for which a significant number of proxy data and modelling studies are now available..."
Line 182: Ought to note that forest expansion (not necessarily evergreen – possibly larch dominant in at least some regions), agrees with proxy data. See, for example, this review of Arctic NA Neogene flora
Fletcher, T.L., Telka, A., Rybczynski, N., Matthews, J.V., Jr, 2021. Neogene and early Pleistocene flora from Alaska and Arctic/Subarctic Canada: New data, intercontinental comparisons and correlations. Palaeontol Electron 24, a07. https://doi.org/10.26879/1121Line 192: Also https://doi.org/10.1130/2021.2551(01)
Line 207: Useful articles overlooked here, as per 1, above.
Line 227: Useful articles overlooked here, as per 1, above.
Line 356: This contradicts the 470 ppm CO2 for the Early Pliocene cited at line 164.
Line 362: Presumably MIO also has several other important differences in boundary conditions. Note the main ones?
Line 388: Any reason why? Also, how does this study differ from Lohmann et al. 2022.?
Line 403: Was this chosen as the models appeared to be at equilibrium, because it is a standard used this this model, or was this a more pragmatic decision?
Line 564: This section opens with uncertainty, and then seems to describe a quite certain timeline. Are parts of this settled and other parts not?
Line 653: Relies throughout on the reader having previously read Lohmann et al 2022. While, of course, not every details should be repeated, it seems some key information should be included that explains the authors choices: Why use only the lowest value from Lohmann et al 2022? Is this the most like according to observations or mechanistic modelling?
Figure 8: See above discussion of missing potential terrestrial proxy sites for DMC.
Figure 12 caption Line 882: the use of the term "implied" for ocean meridional heat transport is unclear. As described in Section 3.3, both the total and oceanic transports are diagnostically inferred from integrated fluxes rather than directly calculated from model transport fields. Consider clarifying the terminology in the caption and figure legend, or using consistent terminology for all inferred transport quantities.
Line 925: Directly relevant paper:
Feng, R., Otto‐Bliesner, B.L., Xu, Y., Brady, E., Fletcher, T., Ballantyne, A., 2019. Contributions of aerosol‐cloud interactions to mid‐Piacenzian seasonally sea ice‐free Arctic Ocean. Geophys. Res. Lett. 46, 9920–9929. https://doi.org/10.1029/2019gl083960
Line 937: The discussion ignores the seasonal distribution of cloud effects and misses a literature connection to winter Arctic warmth and seasonal data-model mismatch
Tindall, J., Haywood, A., Salzmann, U., Dolan, A., Fletcher, T., 2022. The Warm Winter Paradox in the Pliocene Northern High Latitudes. Clim. Past 18, 1385–1405. https://doi.org/10.5194/cp-18-1385-2022
Line 947: Add more recent citation given you are saying still but referring to a 13 year old article?
Technical Corrections
The writing contains many sentences with multiple nested clauses. It would be easier for the reader if these were split into separate sentences. Examples include the sentences starting at Line 52; Line 68, and Line 78. This also obfuscates which citation supports which claim.
There are many other instances throughout that would also benefit from additional editing for English expression.
Line 122: Mixed use of COSMOS with and without the definite article. Remove “the” before COSMOS?
Line 207: Duplication. Csank et al 2011c cited here, is the same article as Csank et al 2011b.
Figure 2 caption, Lines 209-223 : The caption refers to ODP Site 911 IRD, ODP Site 910 IP25, and ODP Site 907 IP25. A reader would benefit from explicitly identifying which plotted symbol corresponds to each site in the figure itself or caption.
Line 242: Independent of?
Line 329: Otto-Bliesner et al., 2017 is not in the reference list.
See also:
Feng, R., Otto-Bliesner, B.L., Fletcher, T.L., Tabor, C.R., Ballantyne, A.P., Brady, E.C., 2017. Amplified Late Pliocene terrestrial warmth in northern high latitudes from greater radiative forcing and closed Arctic Ocean gateways. Earth Planet. Sci. Lett. 466, 129-138. https://doi.org/10.1016/j.epsl.2017.03.006
Line 331: This sentence reads as though Arctic feedbacks or polar amplification were explicitly suppressed in a dedicated model experiment. However, Salzmann et al. (2013) identified differences in the strength of Arctic feedbacks among PlioMIP models and found that models exhibiting weaker Arctic feedbacks generally failed to simulate ice-free Arctic summers. Consider rephrasing for clarity.
Line 334: Needs clarification,
Line 412: (CDO)? Misplaced?
Line 449: Define SIC (Sea Ice Concentration) in the text.
Line 466: Remove the duplicated full stop after "region of interest" ("interest. .").
Figure 4 caption, line 586: "2σ-treshold" → "2σ-threshold".
Sites could do with labels as the text refers to regions using the sites as landmarks.
Labels above each map in the figure (and 5 and 6) are confusing. Is it the COSMOS output shown on the blue scale? And the difference only the region of disagreement. It doesn’t seem to be a quantitative map of the difference in sea ice between COSMOS and CMIP6, for example.
What is CMIP3 CMIP5 PI ensemble mean. How is CMIP3 involved?
Stippling for disagreement and shading l for land sea-mask changes isn’t very clear. Lines for sea ice boarders were also difficult to see. More contrasting colour may help.
Line 505: "minorly impacted" is unusual scientific wording; consider "only weakly impacted" or "only weakly affected".
Line 533-536: Subject-verb agreement: "the COSMOS largely simulate" → "the COSMOS largely simulates" (or define COSMOS consistently as plural throughout).
Line 460: "The COSMOS have been extensively applied..." should be made consistent with usage elsewhere. The manuscript alternates between treating COSMOS as singular and plural.
Line 515: "the COSMOS simulate" should be checked for consistency with the chosen singular/plural treatment of COSMOS.
Line 533: "the COSMOS are a bit colder" should likewise be made consistent.
Figure 8 Caption: Line 710: Is MJJAS as per this label, this is the mollusc paper Csank et al 2011b, not the JJ reconstruction from tree ring isotopes in Csank et al 2011a.
Line: 722 “bar, proxy, or model.” typo
Line 730: "annual pre-industrial SIC (50-90 %)" is described as annual while the following discussion concerns the seasonal cycle. Clarification may be needed.
Figure 9 caption, line 762: "0°C iso-therm" → "0°C isotherm".
line 762: Grey line for 0 m iso-thickness not visible as it is the x-axis.
Line 766: fix brackets
Figure 10 caption,
Line 802: insert white. (white, brown, …) to align with text later in the figure caption.
line 810: Extra opening parenthesis in "(Climate Data Operators; (Schulzweida, 2023)."
Line 825: Spanning here could be misleading. Bracketing?
Line 844: Exceptions are
Line 902: Should Kneis et al 2014a also be cited here?
Line 904: Suggests rather than implies? I don’t think implies is what you mean, exactly.
Line 1004: "interal wave generation" → "internal wave generation".
Line 1030: "wave-braking" should likely be "wave breaking".
Line 1124: Author initials are inconsistent: "S.d.S" vs elsewhere "S.D.S."
Line 1110-1112: Formatting of author initials is inconsistent ("S.T.B.:", "S.D.S.:", "G.L.:"). Check punctuation placement.
Reference list: "Mcclymont" and "McClymont" both appear in the manuscript text; capitalisation should be checked for consistency throughout. Examples include lines 96, 185, and the reference section.
Reference list: "Nature communications", "Scientific reports", "Geophysical research letters", "Ocean modelling", etc. appear with inconsistent journal-title capitalisation.