the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 24 Feb 2026
Simulation of climate during the Miocene Climatic Optimum under different CO2 forcings
Abstract. The Miocene Climatic Optimum (MCO; 17 – 14 Ma), characterized by global mean surface temperatures ~7 °C higher than preindustrial (PI), offers a target for validating models for warmer-than-present climate states. Here, we use a water isotope tracer enabled version of the Community Earth System Model to simulate the MCO under 1x (MCO1x), 2x (MCO2x), and 4x (MCO4x) PI CO2. Our simulations show significant warming due to MCO boundary conditions as well as a small increase in equilibrium climate sensitivity with higher CO2. All simulations exhibit a decreased mean equator-to-pole temperature gradient relative to PI. However, the spatial patterns of warming are distinct between simulations with relatively greater high latitude warming between MCO1x and MCO2x and relatively greater low latitude warming between MCO2x and MCO4x. Warming is associated with enhanced precipitation and enriched precipitation δ¹⁸O (δ¹⁸Op) globally.
We compare the MCO model outputs with proxy of surface temperature, precipitation, and δ¹⁸Op. Like many other MCO modeling studies, our simulations underestimate the reduced latitudinal temperature gradient reconstructed with proxies. We find better model-proxy agreement for terrestrial and marine temperature records in the MCO1x and MCO4x experiments, respectively. Precipitation and δ¹⁸Op records show the best agreement with the MCO2x and MCO1x experiments, respectively, but there are large uncertainties due to limited proxy data and large reconstruction uncertainties. The MCO2x simulation is warmer than the projected 2080–2100 climate under RCP8.5, highlighting the importance of both boundary conditions and equilibrium versus transient climate system response to increased CO2.
Competing interests: I declare that neither I nor my co-authors have any competing interests.
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.- Preprint
(3273 KB) - Metadata XML
-
Supplement
(3009 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on egusphere-2026-899', Anonymous Referee #1, 07 Mar 2026
-
AC1: 'Reply on RC1', Hamida Nadoya, 31 Aug 2026
Publisher’s note: this comment is a copy of AC3 and its content was therefore removed on 2 September 2026.
Citation: https://doi.org/10.5194/egusphere-2026-899-AC1 -
AC3: 'Reply on RC1', Hamida Nadoya, 31 Aug 2026
We thank the reviewer for their thoughtful comments. We have responded to all comments below and provided an updated manuscript with tracked changes. Below, original reviewer comments are red, our responses are plain text, and manuscript quotes are blue.
Major comments
1. Please clarify what is genuinely new about this set of Miocene iCESM1.2 simulations relative to prior studies using the same model (e.g., Acosta et al., Paleoceanogr. Paleoclimatol., e2021PA004383; Liu et al., Geophys. Res. Lett., e2024GL109159). As presented, the primary differences seem to be in the paleogeography, vegetation, and ice-sheet boundary conditions. If so, state this explicitly and explain why these choices are critical for the simulated climate response—for instance, how altered topography or gateways influence circulation patterns, or how vegetation/ice changes affect albedo and feedbacks.
Yes, boundary conditions and CO2 are major differences between studies. The two studies mentioned (Acosta et al., 2022; Liu et al., 2024) used Herold et al. (2011) boundary conditions and CO2 of 400 ppm. However, their model simulations could not capture the proxy reconstructed temperature during the MCO, thus we use higher CO2 of 2x and 4x PI level. Also, neither study included water isotope tracers; Acosta et al. (2024) used CLM4.5 for the land model; Liu et al. (2024) used CAM4 for the atmosphere model. We have also added the following text to clarify the experiment differences.
“Many previous modeling studies of the MCO used the Herold et al. (2008) boundary conditions (Krapp and Jungclaus, 2011; Burls et al. 2021; Herold et al., 2010; Liu et al., 2024; Tan et al., 2026). Although Sun et al. (2024) used Frigola et al. (2018) boundary conditions, they worked with a different climate model than the one used in our study. Our study is the first to use iCESM with Frigola et al. (2018) boundary conditions. Different climate models and boundary conditions can lead to large differences in MCO simulations. For example, Hutchinson et al. (2025) explored the role of different paleogeographic configurations and found that a prominent Greenland Scotland ridge favors a stronger Atlantic Meridional Overturning Circulation (AMOC) whereas an open Arctic-Atlantic connections slows down AMOC and activates the Pacific Meridional Overturning Circulation (PMOC). Furthermore, Liu et al. (2024) reported that an open Canadian Arctic Archipelago allows for North Atlantic deep water formation and strengthening of the AMOC. In addition, initial ocean conditions may play a role in simulated climate. Lee et al. (2010) found that initialization from a cold climate state resulted into a stronger AMOC whereas initialization from a warm climate state resulted into a weaker AMOC. Thus, small changes in boundary conditions and initial conditions can have a large influence on climate.”
I recommend a concise table (or bulleted summary) comparing your boundary conditions to standard datasets like Herold et al. (2008), covering: land–sea mask/topography, ocean gateways/bathymetry, prescribed vegetation/land surface, ice-sheet extent/height, CO₂ and other greenhouse gases, aerosols, and orbital parameters. This would enable readers to attribute climate differences (e.g., in heat transport or precipitation) to boundary conditions versus intrinsic model behavior.
We have added the following text and table.
“In comparison to Herold et al. (2008) boundary conditions, Frigola et al. (2018) boundary conditions include higher sea level of ~4 m, a closed Tethys seaway, a narrower Central American Seaway, shallower Indonesian throughflow barriers, and an Antarctic ice sheet extent of ~6 km3.”
Table 1. Comparison between Frigola et al. (2018) and Herold et al. (2008) MCO boundary conditions and the preindustrial.
Preindustrial
Frigola et al. (2018)
Herold et al. (2008)
Topography
Present day
Modified Herold et al. (2008) with changes in tropical seaways and Antarctic ice sheet
Original Herold et al. (2008)
Ocean gateways
Closed Tethys, Closed Central American Seaway, Constrained Indonesian throughflow
Open Tethys, Narrower Central American, Shallower Indonesian throughflow
Closed Tethys, Wider Central American Seaway, Deeper Indonesian throughflow
Ice sheets
Ice in both northern and southern hemisphere, include 2.9 million km3 of ice on Greenland and 28.5 million km3 of ice on Antarctica
Ice free northern hemisphere, 6 million km3 of ice on Antarctica
Ice free northern hemisphere, 16 million km3 of ice on Antarctica
Sea level
Present day
48 m higher than PI
52 m higher than PI
Vegetation
Present day
Combination of Pound et al. (2012), Wolfe (1985), Morley (2011)
N/A
Antarctica topography
Present day
Pollard and DeConto (2012)
Reduced by 1000 m
Importantly, for any differences in these simulation results compared with previous iCESM1.2 simulations (e.g., temperatures, hydroclimate, or δ¹⁸O), provide mechanistic explanations. Link these to physical processes such as changes in boundary / initial conditions where possible.
It can be difficult to determine the exact drivers of the differences between MCO simulations. We have added comparisons in the results sections. We primarily compared differences in surface temperatures which most previous studies focused on. Below is the text we have added.
“In contrast, similar magnitudes of warming are observed in past studies that used CESM1.2 such as Lee et al. (2025) with 18.7°C and 20.0°C global mean surface temperatures at 1x PI and 400 ppm CO2 levels, respectively.
“In the tropics, the mean position of the Intertropical Convergence Zone (ITCZ) is in the SH with maximum annual mean precipitation exceeding 6 mm/day, which is consistent with previous Miocene simulations (Acosta et al., 2023) and likely a result of shift in the hemispheric energy balance (Donohoe et al., 2013).”
2. From a MioMIP perspective, please situate your iCESM1.2 ensemble relative to other MioMIP model results available at https://www.deepmip.org/data-miocene/. Are the large-scale Miocene patterns (e.g., polar amplification, tropical SST gradients, monsoon intensity, zonal mean precipitation shifts, sea-ice extent) broadly consistent, and where does iCESM deviate? Identify explicitly what this study adds beyond existing MioMIP findings—such as new water-isotope diagnostics, refined boundary conditions, a unique climate mechanism, or improved process attribution (e.g., via energy balance, heat transport, or cloud feedbacks).
We have compared our simulations with the available MioMIP model data at 1x and 2x PI CO2. Many simulations used 400 ppm and 3x PI CO2. We compared the zonal mean surface temperature and precipitation and highlighted the similarities and differences and potential attributions. We have added in the following paragraph to our discussion section.
“Compared to available MioMIP1 simulation with 1x and 2x PI CO2, iCESM simulates warmer MCO temperatures (Fig. SP16). Our simulations better capture the proxy reconstructed polar amplification in the NH, though they still underestimate the magnitude of warming. However, in the SH high latitudes, iCESM appears colder than some models such as NorESM-L and HadCM3L-NoICE. Compared to surface temperature, zonal mean precipitation shows large differences between models. Our simulations depict wetter conditions relative to other models in the mid and high latitudes. Tropical precipitation tends to show better agreement between modeling groups with many models simulating a southern hemisphere ITCZ precipitation maximum. The cause of these differences are hard to determine and may be a result of boundary conditions (Naik et al., 2025; Hutchison et al., 2025), climate sensitivity of the models (Burls et al., 2021), initial ocean conditions (Lee et al., 2025), and ability of the model to simulate ocean-ice dynamics (Tan et al., 2026). Notably, Tan et al. (2026) attributed the ability of NorESM to better simulate high latitude temperatures during the MCO to its stronger ocean overturning circulation with enhanced poleward heat and salt transport, resulting in near total sea ice loss.”
3. The current Results section reads primarily as descriptive documentation. The manuscript would be much stronger if the authors emphasized the most exciting outcomes—those that either revise our understanding of Miocene climate or demonstrate new model insights. Identify explicitly which findings are novel and discuss their broader implications.
For the sake of completeness, we describe the experiments in detail in the Results Section. We highlight particularly novel / interesting findings and their implications in the Discussion Section. The topics highlighted in the discussion include; the large warming of 4.1°C due to Miocene boundary conditions compared to previous Miocene studies, the differences in ECS of our simulations at low and high CO2 compared to that of CESM1 and future projections, and the drivers of zonal and regional surface temperature changes. Furthermore, we discuss differences between model-data agreements for surface temperature, precipitation and precipitation d18Oand the implications of which CO2 configuration simulates a better model-proxy data agreement. Lastly, we provided insights on how our simulations compare with available MioMIP simulations and discuss the differences between modeling groups, and implications of the MCO as a future climate analog.
4. Consider merging Section 3.4.1 (“Comparison with preindustrial climate”) and Figure 10 with Section 3.1 and the earlier figures. Since the preindustrial (PI) reference curves already appear in Figure 1, integrating the full PI comparison earlier would make the narrative more cohesive and help readers interpret Miocene–PI anomalies throughout the paper.
“The MCO1x experiment is warmer than the PI experiment with global mean annual temperatures of 18.4°C versus 14.4°C. Therefore, the MCO boundary conditions result into a global warming of 4.0°C, which is generally higher than the warming due to MCO boundary conditions found in previous MCO simulations such as Goldner et al. (2014) with warming of 2.4°C, Krapp and Jungclaus (2012) with warming of 0.7°C, and Burls et al. (2021) with model average warming of 2.0°C. In the NH high-latitudes, MAT is -8.8°C for MCO1x and -18.4°C for PI whereas in the SH high-latitude, MAT is -13.0°C for MCO1x and -31.9°C for PI (Fig. 2a-2b, Fig.2e). The mid-latitude MAT is 11.4°C for MCO1x and 8.9°C for PI in the NH. In the SH, the mid-latitude M AT is 15.6°C and 11.1°C for the MCO1x and PI, respectively. The subtropical MAT is 26.0°C for the MCO1x both in the NH and SH whereas for PI, MAT is 24.2°C in the NH and 23.4°C in the SH. Moreover, MAT in the tropics is also elevated at 28.7°C for the MCO1x relative to 26.7°C for PI. Furthermore, seasonal patterns show pronounced warming over the high-latitudes during winter and over land during summer for MCO1x relative to PI (Fig. SP2; SP10)."
"MCO1x is wetter with global mean annual precipitation of 3.3 mm/day compared to PI with a global mean annual precipitation of 3.0 mm/day (Fig. 2c, d, & h). Over the high-latitudes, average precipitation is 1.4 mm/day for MCO1x and 1.1 mm/day for PI in the NH, whereas in the SH, average precipitation is 1.9 mm/day for MCO1x and 1.3 mm/day for PI. In the NH mid-latitudes, average precipitation is 2.7 mm/day for MCO1x and 2.3 mm/day for PI. On the other hand, SH mid-latitude precipitation averages at 3.2 mm/day for MCO1x and 2.9 mm/day for PI. Moreover, tropical rainfall maximizes at 6.3 mm/day in the SH during the MCO and 6.4 mm/day in the NH for PI. The spatial distribution of precipitation during the MCO relative to PI shows enhanced precipitation over South America and decreased precipitation over eastern and southern Africa. Furthermore, enhanced precipitation is observed over coastal North America, Europe, Eurasia, coastal East Asia, and over the tropics. Over the ocean, enhanced precipitation occurs in the Pacific and Indian Oceans. In contrast, a drier pattern is observed over the central Atlantic and west equatorial Pacific and a large portion of the North Atlantic. In addition, there is expansion of the tropical rain belt in the MCO1x compared to PI, particularly over the equatorial Pacific and Indian Oceans”.
“Global mean annual d18Op is -7.6 ‰for MCO1x and -8.9‰for PI. Previous studies have attributed enrichment in warmer climates to an increase in water vapor residence time (Singh et al., 2016; Winnick et al., 2014, 2015; Kukla et al., 2019). There is also a decrease in the meridional gradient of d18Op during the MCO relative to PI due primarily to polar enrichment (Fig .2). In the NH high-latitudes, average d18Op values are -16.5 ‰ for MCO1x and -17.6‰for PI, and in the SH high-latitudes, average d18Op are -14.7‰ for MCO1x and -21.6‰for PI. The large differences in the high latitude d18Op between simulations likely results from polar amplification of the warming and the resulting temperature effect (Dansgaard, 1964). The midlatitudes exhibit smaller differences in d18Op between the MCO1x and PI climates, with a MCO1x average of -10.2‰ and a PI average of -11.0‰ in the NH and -8.1‰ and -9.1‰ in the SH respectively. MCO1x also shows more enriched in the tropics with an average d18Op of -4.0‰ compared with a PI average d18Op of -5.1 ‰".
"Regionally, enriched d18Op values are observed over the tropics, especially tropical Atlantic, eastern tropical Pacific, and Indian Oceans in MCO1x relative to PI. In addition, MCO1x reveals more positive d18Op values over Africa, particularly North Africa, possibly associated with changes in vegetation cover. On the other hand, the North Atlantic shows depletion in the MCO1x relative to PI, which agrees with the findings of Sun et al. (2024). The lower isotopic values over the North Atlantic in MCO1x may be attributed to the temperature effect as some cooling is observed over the region likely due to an open Central American Seaway, which allows for mixing of the Atlantic and Pacific waters and weakens northward heat transport. The largest difference in d18Op is found over Antarctica, where d18Op is enriched by ~20‰ in MCO1x due to the reduced ice volume, which reduces elevation and albedo and raises temperatures.”
5. Including an Atlantic Meridional Overturning Circulation (AMOC) diagnostic would enhance the discussion of oceanic heat transport. A figure showing the maximum Atlantic overturning stream function (and, if possible, a depth–latitude structure) would connect directly to the meridional heat flux analysis in Figure 7. A concise global overturning depiction could also help link the ocean circulation to the modeled heat transport and energy balance frameworks.
We have added a depth latitude structure of the global, Atlantic, and Indo-Pacific stream function plots for the meridional overturning circulation for the MCO1x experiment and PI in the main text and the differences with MCO2x and MCO4x as supplemental figures. We also added the following text to the results sections:
“Comparing MCO1x with PI, there is less southward ocean heat transport, maximizing at 1.0 PW during the MCO and 1.5 PW for PI over the total global ocean (Fig 7). Large changes in poleward ocean heat transport are observed in the Pacific, with a reduction in southward heat transport in the South Pacific of about 0.5 PW for the MCO1x relative to PI. Moreover, enhanced northward heat transport is observed in the North Pacific during the MCO maximizing at 1.3 PW and 0.8 PW during PI. The reduction in SH ocean heat transport in the Pacific and increase in the NH northward Pacific heat transport suggests a reversal in ocean circulation during the MCO. In the Atlantic, there is less northward heat transport in the North Atlantic with a maximum of 0.2 PW during the MCO compared to 0.8 PW for PI, suggesting a weaker Atlantic Meridional Overturning circulation (AMOC) during the MCO. Moreover, ocean heat transport in the South Atlantic shows a small amount of southward heat transport maximizing at 0.5 PW during the MCO whereas PI shows northward heat transport. The reduction in northward heat transport is further shown in Fig. 9 which shows a weakened AMOC cell for MCO1x (Fig. 9a) compared to a stronger and deeper MOC cell for PI (Fig. 9d). The AMOC cell for MCO1x maximizes at 8Sv while the PI cell maximizes at 16 Sv. This implies decreased deep-water formation in the North Atlantic during the MCO compared to PI. Increasing CO2 from 1x to 2x PI level drives a large intensification of the shallow Indo-Pacific cell whereas increase from 2x to 4x results into minor changes in circulation. The AMOC shows small changes to increasing CO2.”
Minor comments
1. In Tables 1 and 2, the listed values appear overly precise for the intended context. Rounding to single-digit precision (or justifying the current precision) would improve readability and consistency.We have rounded off to one decimal point for all tables.
-
AC1: 'Reply on RC1', Hamida Nadoya, 31 Aug 2026
-
RC2: 'Comment on egusphere-2026-899', Anonymous Referee #2, 13 Apr 2026
This manuscript presents Miocene Climatic Optimum simulations using the water-isotope-enabled CESM under 1x, 2x, and 4x preindustrial CO2, and compares the simulations with proxy constraints on temperature, precipitation, and precipitation isotopes. The topic is important, and the isotope-enabled framework is potentially a valuable addition to Miocene climate research. However, I do not think the current manuscript yet fully capitalizes on the main scientific opportunity provided by this modeling framework, and several of the broader comparative conclusions are presently stronger than the available analysis can support. I therefore recommend major revision.
Major comments
1. The isotope-enabled aspect of the study is the clearest distinguishing feature, but the isotope results are not yet interpreted deeply enough. To me, the main added value of this paper relative to prior Miocene CESM-style studies is the inclusion of water isotopes. However, the manuscript currently presents the isotope results more as an additional diagnostic than as a central scientific contribution. In particular, the paper would be much stronger if it more explicitly discussed the relative roles of temperature effects, precipitation amount effects, moisture source changes, transport pathway shifts, circulation changes, and seasonality in shaping the simulated isotope response. Without this deeper interpretation, the isotope-enabled framework remains underused, despite being one of the manuscript’s main distinguishing features.
2. The comparison with future climate can be informative in a qualitative sense, but the conclusions should be framed more cautiously. I think this comparison can be useful as a broad conceptual contrast, but only if its limitations are made fully explicit. The Miocene and the late 21st century differ fundamentally in paleogeography, land–sea distribution, gateways, vegetation, cryosphere state, and equilibration timescale. These differences strongly limit direct comparison. I therefore encourage the authors to retain this section only as a qualitative discussion and to avoid overly absolute conclusions that might imply stronger comparability than is actually justified.
3. The ranking of different CO2 experiments based mainly on RMSE is not, by itself, the most informative way to interpret the model–proxy comparison. According to the study, terrestrial temperatures agree best with one experiment, marine temperatures with another, precipitation with another, and δ18Op with another, while the manuscript also acknowledges limited data availability and substantial uncertainty. This is an interesting result, but I do not think the main scientific takeaway should simply be that different proxy classes “prefer” different CO2 concentrations based on RMSE. A more meaningful discussion would examine regional pattern agreement and, more importantly, what the mismatches imply about the limitations of the model, the proxy compilations, and the experimental design.
Minor comments
- The current Introduction lacks a clear statement elaborating the novelty and motivation of this study.
- Lines 62–66: The review of previous modeling work is too brief and lacks key references (Herold et al., 2011; Krapp & Jungclaus, 2011; Goldner et al., 2014; Burls et al., 2021 MioMIP1; Acosta et al., 2024; Sun et al., 2024; Liu et al., 2024; Lee et al., 2025; Hutchinson et al., 2025; Tan et al., 2026). It would be valuable to compare different MCO modeling studies and pinpoint the current weaknesses as well as the gap between model results and proxy records.
- The authors mention the climate sensitivity of GCMs but do not elaborate on why it is important for MCO simulations, nor do they discuss the climate sensitivity of the model used in this study.
- Line 90: What is the exact elevation change applied in this study?
- Lines 96–103: What initial conditions (ocean temperature and salinity) were used for the MCO2x simulation? What are the background oxygen isotope values prescribed for each ocean basin, and are they inherited from the PI condition?
- It would be helpful to add the PI results to Figure 1 and Figure 7 for direct comparison.
- Schouten is cited as 2002 in the main text but appears as 2013 in the reference list. The CenCO₂PIP / Hönisch reference is listed twice (Line 630 and Line 633). Only Haywood 2020 is cited in the main text (Line 36), others are not cited but appear in the reference list. Kürschner et al. 2008 is in the reference list but not cited in the main text. Please cross-check all references.
- I would encourage the authors to deposit their data in a public repository rather than making them available only upon request.
Citation: https://doi.org/10.5194/egusphere-2026-899-RC2 -
AC2: 'Reply on RC2', Hamida Nadoya, 31 Aug 2026
We thank the reviewer for their thoughtful comments. We have responded to all comments below and provided an updated manuscript with tracked changes.
Major comments
The isotope-enabled aspect of the study is the clearest distinguishing feature, but the isotope results are not yet interpreted deeply enough. To me, the main added value of this paper relative to prior Miocene CESM-style studies is the inclusion of water isotopes. However, the manuscript currently presents the isotope results more as an additional diagnostic than as a central scientific contribution. In particular, the paper would be much stronger if it more explicitly discussed the relative roles of temperature effects, precipitation amount effects, moisture source changes, transport pathway shifts, circulation changes, and seasonality in shaping the simulated isotope response. Without this deeper interpretation, the isotope-enabled framework remains underused, despite being one of the manuscript’s main distinguishing features.
We acknowledge the value of an in-depth analysis and discussion of the stable water isotopes in our MCO experiments. However, a detailed analysis of the d18Op responses is outside the scope of the current study. We plan to analyze the isotopic response in greater detail in future works. We will also make the model data publicly available in case others want to explore the isotopic responses in greater detail. That said, we have added a comparison of simulated d18Op for MCO1x and PI experiments as well as a brief discussion of the mechanisms to the main text as shown below.
“Global mean annual d18Op is -7.6 ‰for MCO1x and -8.9‰for PI. Previous studies have attributed enrichment in warmer climates to an increase in water vapor residence time (Singh et al., 2016; Winnick et al., 2014, 2015; Kukla et al., 2019). There is also a decrease in the meridional gradient of d18Op during the MCO relative to PI due primarily to polar enrichment (Fig .2). In the NH high-latitudes, average d18Op values are -17.0 ‰ for MCO1x and -18.5‰for PI, and in the SH high-latitudes, average d18Op are -17.3‰ for MCO1x and -27.9‰for PI. The large differences in the high latitude d18Op between simulations likely results from polar amplification of the warming and the resulting temperature effect (Dansgaard, 1964). The midlatitudes exhibit smaller differences in d18Op between the MCO1x and PI climates, with a MCO1x average of -9.5‰ and a PI average of -10.4‰. MCO1x also shows more enriched in the tropics with an average d18Op of -4.0‰ compared with a PI average d18Op of -5.1 ‰.
"Regionally, enriched d18Op values are observed over the tropics, especially tropical Atlantic, eastern tropical Pacific, and Indian Oceans in MCO1x relative to PI. In addition, MCO1x reveals more positive d18Op values over Africa, particularly North Africa, possibly associated with changes in vegetation cover. On the other hand, the North Atlantic shows depletion in the MCO1x relative to PI, which agrees with the findings of Sun et al. (2024). The lower isotopic values over the North Atlantic in MCO1x may be attributed to the temperature effect as some cooling is observed over the region likely due to an open Central American Seaway, which allows for mixing of the Atlantic and Pacific waters and weakens northward heat transport. The largest difference in d18Op is found over Antarctica, where d18Op is enriched by ~20‰ in MCO1x due to the reduced ice volume, which reduces elevation and albedo and raises temperatures.”
To briefly discuss the role of temperature and precipitation effects, we plotted the relationships between model simulated d18Op, temperature, and precipitation (Fig. SP6). We compare annual climatology at each grid cell as these data is most likely reflected in proxy archives of the MCO. We also added the following text.
“Relationships between d18Op, surface temperature, and precipitation are generally consistent with increasing CO2 in our MCO simulations. As expected, mean annual surface temperatures show a strong positive relationship with mean annual d18Op outside of the tropics for the three MCO experiments (Fig. SP6; Table SP1). MCO1x exhibits the most depleted d18Op on average and MCO4x exhibits most enriched d18Op on average. The d18Op-temperature slopes also shift slightly with surface warming, but the differences are minimal. There is also a weakly positive relationship between d18Op and temperature and warming outside of the tropics in the MCO simulations. In contrast, the surface temperature shows no relationship with d18Op in the tropics for all the three MCO experiments. Likewise, precipitation outside the tropics does not show a statistically significant relationship with d18Op. The precipitation amount effect is only well observed in the tropics (Fig. SP6b) where increase in precipitation shows more depleted d18Op; the negative correlation between d18Op and precipitation is enhanced with warming."
"Our simulations suggest that on long time and spatial scales, d18Op in the MCO may be loosely used as a proxy for surface temperature in the mid-to-high latitudes and precipitation in the tropics. However, the correlations may not explain the underlying mechanisms. A recent study by Siler et al. (2025) suggests that d18Op is highly influenced by precipitation and distance from the source as it largely varies by how much evaporation and transport distance change upstream. Their study explains the amount effect as resulting from evaporation being largely uniform in the tropics and thus variability in d18Op is influenced by the transport distance, which is inversely proportional to precipitation. In the mid-to-high latitudes, the observed temperature effect is driven by gradients in evaporation, which depend largely on surface temperature. A more detailed study of the underlying mechanisms for d18Op is necessary to determine if these relationships hold under MCO boundary conditions.”
Table SP1. Coefficient of determination (R2) showing d18Op-temperature and precipitation relationships
Temperature
Precipitation
MCO1x (15-90°)
0.93
0.00
MCO2x (15-90°)
0.92
0.00
MCO4x (15-90°)
0.89
0.00
MCO1x(Tropics)
0.07
0.38
MCO2x(Tropics)
0.09
0.35
MCO4x(Tropics)
0.10
0.41
MCO2x-MCO1x (15-90°)
0.46
0.10
MCO4x-MCO2x (15-90°)
0.06
0.01
MCO2x-MCO1x (Tropi.cs)
0.01
0.17
MCO4x-MCO2x (Tropics)
0.00
0.19
2. The comparison with future climate can be informative in a qualitative sense, but the conclusions should be framed more cautiously. I think this comparison can be useful as a broad conceptual contrast, but only if its limitations are made fully explicit. The Miocene and the late 21st century differ fundamentally in paleogeography, land–sea distribution, gateways, vegetation, cryosphere state, and equilibration timescale. These differences strongly limit direct comparison. I therefore encourage the authors to retain this section only as a qualitative discussion and to avoid overly absolute conclusions that might imply stronger comparability than is actually justified.
We acknowledge the limitations of comparing the MCO with present day. We attempted to provide the most direct comparison possible by chosen simulations with similar amounts of global warming i.e MCO2x and MCO1x and RCP8.5 and PI. To further point out the limitations of comparing the MCO with future climate change, we added in the following statement before presenting the comparison.
“There exists no perfect near future climate analog as past climates have different boundary conditions compared to present day. Moreover, there is uncertainty surrounding the boundary conditions of past climates as well as the past climate reconstructions themselves. Although the MCO is characterized by only slightly different paleogeography compared to present day, it also has differences in vegetation and land ice. These differences can alter climate sensitivity due to changes in feedbacks. Furthermore, the MCO is assumed to be in an equilibrium climate state while RCP8.5 represents ongoing warming. Thus, while we compare climate responses between the MCO1x and MCO2x and projected changes under RCP8.5 relative to PI, the relationship should be considered qualitative."
3. The ranking of different CO2 experiments based mainly on RMSE is not, by itself, the most informative way to interpret the model–proxy comparison. According to the study, terrestrial temperatures agree best with one experiment, marine temperatures with another, precipitation with another, and δ18Op with another, while the manuscript also acknowledges limited data availability and substantial uncertainty. This is an interesting result, but I do not think the main scientific takeaway should simply be that different proxy classes “prefer” different CO2 concentrations based on RMSE. A more meaningful discussion would examine regional pattern agreement and, more importantly, what the mismatches imply about the limitations of the model, the proxy compilations, and the experimental design.
Comparison with proxy reconstructions and selecting out the best CO2 is such an important aspect of the study. We agree detailed regional comparison is necessary for a better ranking, but it is challenging with the limited data available and the spatial distribution of the data. We have, however, added a few comparisons and modified the overall conclusion on the best CO2 configurations. We have added the following.
“Regional comparisons generally show better model-proxy agreement between SSTs and MCO2x at sites in the tropical oceans except those in the North Atlantic and two sites in the Southern Ocean. Available high-latitude proxy reconstructions for SST are mostly over the North Atlantic and simulated MCO surface temperatures are too cold in this region, which could partly be explained by the cold bias in present-day CESM1 simulations (Park et al., 2014).”
“In addition, most reconstructed terrestrial temperatures are located in high elevation areas. Our simulations use a relatively low-resolution configuration of CESM. Although this is a standard resolution for paleoclimate simulations, it does result in smoothed topography, which could contribute to model biases in the simulated climate. Furthermore, paleo-elevations uncertainties are typically greater for high elevation sites. Finally, the terrestrial temperature compilation consists of several proxy reconstruction techniques that involve different assumptions and uncertainties.”
“However, it is different to compare d18Op reconstructions with the MCO simulations because the proxy data are in a single latitudinal band in the NH, hindering a more complete spatial comparison.”
Minor comments
1. The current Introduction lacks a clear statement elaborating the novelty and motivation of this study.
We have added the following statements on the value of simulating water isotopes.
“Moreover, water isotopes provide more robust information about large scale circulation compared to physical variables that are highly influenced by regional processes (Hu et al., 2018). Specifically, stable water isotopes of oxygen help track moisture sources and sinks and reconstruct temperature and ice volume (Kochhann et al., 2017; Holbourn et al., 2014; Miller et al., 1987). Many paleoclimate reconstructions are based on stable water isotopic signatures (Pound et al., 2012; Tripati et al., 2015; Müller, et al., 1998). Thus, model simulations with stable isotopes which allow for direct comparison of proxies and model outputs are helpful for minimizing biases.”
2. Lines 62–66: The review of previous modeling work is too brief and lacks key references (Herold et al., 2011; Krapp & Jungclaus, 2011; Goldner et al., 2014; Burls et al., 2021 MioMIP1; Acosta et al., 2024; Sun et al., 2024; Liu et al., 2024; Lee et al., 2025; Hutchinson et al., 2025; Tan et al., 2026). It would be valuable to compare different MCO modeling studies and pinpoint the current weaknesses as well as the gap between model results and proxy records.
We have rephrased the paragraph. It now reads:
“MCO CO2 reconstructions remain uncertain. Early estimates suggested CO2 levels of less than 500 ppm (Zhang et al., 2013; Greenop et al., 2014). However, MCO simulations with CO2 levels less than 500 ppm underestimate proxy-reconstructed warming, the equator-to-pole temperature gradient, and precipitation (Burls et al., 2021; Liu et al.,2024; Acosta et al., 2022; Lee et al., 2025; Tan et al., 2026). For instance, Goldner et al. (2014) used CESM 1.0 with a slab ocean and atmospheric CO2 of 400 ppm and simulated a global warming of ~4°C relative to PI, which is still ~4°C colder than their proxy reconstructed GMAT. More recently, boron isotope-based estimates suggest CO2 as high as 4x PI (Rae et al., 2021; Hönisch et al., 2023). With higher prescribed CO2, MCO simulations tend to better agree with high-latitude temperatures from proxy reconstructions but may overestimate tropical surface temperatures and the magnitude varies between models (Burls et al., 2021).”
“Many previous modeling studies of the MCO used the Herold et al. (2008) boundary conditions (Krapp and Jungclaus, 2011; Burls et al. 2021; Herold et al., 2010; Liu et al., 2024; Tan et al., 2026). Although Sun et al. (2024) used Frigola et al. (2018) boundary conditions, they worked with a different climate model than the one used in our study. Our study is the first to use iCESM with Frigola et al. (2018) boundary conditions. Different climate models and boundary conditions can lead to large differences in MCO simulations. For example, Hutchinson et al. (2025) explored the role of different paleogeographic configurations and found that a prominent Greenland Scotland ridge favors a stronger Atlantic Meridional Overturning Circulation (AMOC) whereas an open Arctic-Atlantic connections slows down AMOC and activates the Pacific Meridional Overturning Circulation (PMOC). Furthermore, Liu et al. (2024) reported that an open Canadian Arctic Archipelago allows for North Atlantic deep water formation and strengthening of the AMOC. In addition, initial ocean conditions may play a role in simulated climate. Lee et al. (2010) found that initialization from a cold climate state resulted into a stronger AMOC whereas initialization from a warm climate state resulted into a weaker AMOC. Thus, small changes in boundary conditions and initial conditions can have a large influence on climate.”
3. The authors mention the climate sensitivity of GCMs but do not elaborate on why it is important for MCO simulations, nor do they discuss the climate sensitivity of the model used in this study.
We have added the following statements.
“The equilibrium climate sensitivity (ECS) of an Earth system model plays a critical role in determining the magnitude of warming and associated changes in climate under increasing atmospheric CO2 (Sherwood et al., 2020; IPCC, 2021). Paleoclimate reconstructions suggest that ECS may be state dependent with warmer climates such as the MCO exhibiting stronger feedbacks and potentially higher ECS than present-day conditions (Anagnostou et al. (2020); IPCC, 2021)”.
“Our chosen Earth system model has a moderately high equilibrium climate sensitivity and has shown to be one of the best performing models for simulation of the Eocene (Zhu et al., 2019) and Pliocene (Feng et al., 2020)”.
4. Line 90: What is the exact elevation change applied in this study?
The elevation was increased by roughly 1 km over the central Andes between 14 - 27oS following a study by Boschman (2021). We have added this text into the description.
5. Lines 96–103: What initial conditions (ocean temperature and salinity) were used for the MCO2x simulation? What are the background oxygen isotope values prescribed for each ocean basin, and are they inherited from the PI condition?
The initial d18O was set to zero and initial ocean temperature and salinity was taken from a previously spun up 2xCO2 CESM2 Miocene case. We have added this information into the manuscript.
6. It would be helpful to add the PI results to Figure 1 and Figure 7 for direct comparison.
“The MCO1x experiment is warmer than the PI experiment with global mean annual temperatures of 18.4°C versus 14.4°C. Therefore, the MCO boundary conditions result into a global warming of 4.0°C, which is generally higher than the warming due to MCO boundary conditions found in previous MCO simulations such as Goldner et al. (2014) with warming of 2.4°C, Krapp and Jungclaus (2012) with warming of 0.7°C, and Burls et al. (2021) with model average warming of 2.0°C. In the NH high-latitudes, MAT is -8.8°C for MCO1x and -18.4°C for PI whereas in the SH high-latitude, MAT is -13.0°C for MCO1x and -31.9°C for PI (Fig. 2a-2b, Fig.2e). The mid-latitude MAT is 11.4°C for MCO1x and 8.9°C for PI in the NH. In the SH, the mid-latitude MAT is 15.6°C and 11.1°C for the MCO1x and PI, respectively. The subtropical MAT is 26.0°C for the MCO1x both in the NH and SH whereas for PI, MAT is 24.2°C in the NH and 23.4°C in the SH. Moreover, MAT in the tropics is also elevated at 28.7°C for the MCO1x relative to 26.7°C for PI. Furthermore, seasonal patterns show pronounced warming over the high-latitudes during winter and over land during summer for MCO1x relative to PI (Fig. SP2; SP10)."
"MCO1x is wetter with global mean annual precipitation of 3.3 mm/day compared to PI with a global mean annual precipitation of 3.0 mm/day (Fig. 2c, d, & h). Over the high-latitudes, average precipitation is 1.4 mm/day for MCO1x and 1.1 mm/day for PI in the NH, whereas in the SH, average precipitation is 1.9 mm/day for MCO1x and 1.3 mm/day for PI. In the NH mid-latitudes, average precipitation is 2.7 mm/day for MCO1x and 2.3 mm/day for PI. On the other hand, SH mid-latitude precipitation averages at 3.2 mm/day for MCO1x and 2.9 mm/day for PI. Moreover, tropical rainfall maximizes at 6.3 mm/day in the SH during the MCO and 6.4 mm/day in the NH for PI. The spatial distribution of precipitation during the MCO relative to PI shows enhanced precipitation over South America and decreased precipitation over eastern and southern Africa. Furthermore, enhanced precipitation is observed over coastal North America, Europe, Eurasia, coastal East Asia, and over the tropics. Over the ocean, enhanced precipitation occurs in the Pacific and Indian Oceans. In contrast, a drier pattern is observed over the central Atlantic and west equatorial Pacific and a large portion of the North Atlantic. In addition, there is expansion of the tropical rain belt in the MCO1x compared to PI, particularly over the equatorial Pacific and Indian Oceans”.
“Global mean annual d18Op is -7.6 ‰for MCO1x and -8.9‰for PI. Previous studies have attributed enrichment in warmer climates to an increase in water vapor residence time (Singh et al., 2016; Winnick et al., 2014, 2015; Kukla et al., 2019). There is also a decrease in the meridional gradient of d18Op during the MCO relative to PI due primarily to polar enrichment (Fig .2). In the NH high-latitudes, average d18Op values are -16.5 ‰ for MCO1x and -17.6‰for PI, and in the SH high-latitudes, average d18Op are -14.7‰ for MCO1x and -21.6‰for PI. The large differences in the high latitude d18Op between simulations likely results from polar amplification of the warming and the resulting temperature effect (Dansgaard, 1964). The midlatitudes exhibit smaller differences in d18Op between the MCO1x and PI climates, with a MCO1x average of -10.2‰ and a PI average of -11.0‰ in the NH and -8.1‰ and -9.1‰ in the SH respectively. MCO1x also shows more enriched in the tropics with an average d18Op of -4.0‰ compared with a PI average d18Op of -5.1 ‰.
“Relationships between d18Op, surface temperature, and precipitation are generally consistent with increasing CO2 in our MCO simulations. As expected, mean annual surface temperatures show a strong positive relationship with mean annual d18Op outside of the tropics for the three MCO experiments (Fig. SP6; Table SP1). MCO1x exhibits the most depleted d18Op on average and MCO4x exhibits most enriched d18Op on average. The d18Op-temperature slopes also shift slightly with surface warming, but the differences are minimal. There is also a weakly positive relationship between d18Op and temperature and warming outside of the tropics in the MCO simulations. In contrast, the surface temperature shows no relationship with d18Op in the tropics for all the three MCO experiments. Likewise, precipitation outside the tropics does not show a statistically significant relationship with d18Op. The precipitation amount effect is only well observed in the tropics (Fig. SP6b) where increase in precipitation shows more depleted d18Op; the negative correlation between d18Op and precipitation is enhanced with warming."
Our simulations suggest that on long time and spatial scales, d18Op in the MCO may be loosely used as a proxy for surface temperature in the mid-to-high latitudes and precipitation in the tropics. However, the correlations may not explain the underlying mechanisms. A recent study by Siler et al. (2025) suggests that d18Op is highly influenced by precipitation and distance from the source as it largely varies by how much evaporation and transport distance change upstream. Their study explains the amount effect as resulting from evaporation being largely uniform in the tropics and thus variability in d18Op is influenced by the transport distance, which is inversely proportional to precipitation. In the mid-to-high latitudes, the observed temperature effect is driven by gradients in evaporation, which depend largely on surface temperature. A more detailed study of the underlying mechanisms for d18Op is necessary to determine if these relationships hold under MCO boundary conditions.”
“Comparing MCO1x with PI, there is less southward ocean heat transport, maximizing at 1.0 PW during the MCO and 1.5 PW for PI over the total global ocean (Fig 7). Large changes in poleward ocean heat transport are observed in the Pacific, with a reduction in southward heat transport in the South Pacific of about 0.5 PW for the MCO1x relative to PI. Moreover, enhanced northward heat transport is observed in the North Pacific during the MCO maximizing at 1.3 PW and 0.8 PW during PI. The reduction in SH ocean heat transport in the Pacific and increase in the NH northward Pacific heat transport suggests a reversal in ocean circulation during the MCO. In the Atlantic, there is less northward heat transport in the North Atlantic with a maximum of 0.2 PW during the MCO compared to 0.8 PW for PI, suggesting a weaker Atlantic Meridional Overturning circulation (AMOC) during the MCO. Moreover, ocean heat transport in the South Atlantic shows a small amount of southward heat transport maximizing at 0.5 PW during the MCO whereas PI shows northward heat transport. The reduction in northward heat transport is further observed in Fig. 9 which shows a weakened AMOC cell for MCO1x (Fig. 9a) compared to a stronger and deeper MOC cell for PI (Fig. 9d). The AMOC cell for MCO1x maximizes at 8Sv while the PI cell maximizes at 16 Sv. This implies decreased deep-water formation in the North Atlantic during the MCO compared to PI. Increasing CO2 from 1x to 2x PI level drives a large intensification of the shallow Indo-Pacific cell whereas increase from 2x to 4x results into minor changes in circulation. The AMOC shows small changes to increasing CO2.”
7. Schouten is cited as 2002 in the main text but appears as 2013 in the reference list. The CenCO₂PIP / Hönisch reference is listed twice (Line 630 and Line 633). Only Haywood 2020 is cited in the main text (Line 36), others are not cited but appear in the reference list. Kürschner et al. 2008 is in the reference list but not cited in the main text. Please cross-check all references.
We have rectified these issues.
8. I would encourage the authors to deposit their data in a public repository rather than making them available only upon request.
A repository has been created with all the available datasets.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 1,065 | 815 | 117 | 1,997 | 197 | 79 | 139 |
- HTML: 1,065
- PDF: 815
- XML: 117
- Total: 1,997
- Supplement: 197
- BibTeX: 79
- EndNote: 139
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
This manuscript presents new Miocene climate simulations using the water isotope-enabled Community Earth System Model (iCESM 1.2). The authors examine large-scale climate features, including temperature, precipitation, precipitation δ¹⁸O, and meridional heat transport, and provide an energy-balance interpretation of the simulated temperatures. Model results are compared against available proxy records for evaluation.
Overall, the manuscript is clearly written and well structured. The scope of the study aligns well with the readership of Climate of the Past. However, I found it challenging to identify the key novel elements and the specific advances relative to prior Miocene simulations. I encourage the authors to articulate more clearly what is new or distinctive in this work and why those aspects are significant for advancing Miocene modeling. Below are several detailed comments and suggestions.
Major comments
1. Please clarify what is genuinely new about this set of Miocene iCESM1.2 simulations relative to prior studies using the same model (e.g., Acosta et al., Paleoceanogr. Paleoclimatol., e2021PA004383; Liu et al., Geophys. Res. Lett., e2024GL109159). As presented, the primary differences seem to be in the paleogeography, vegetation, and ice-sheet boundary conditions. If so, state this explicitly and explain why these choices are critical for the simulated climate response—for instance, how altered topography or gateways influence circulation patterns, or how vegetation/ice changes affect albedo and feedbacks.
I recommend a concise table (or bulleted summary) comparing your boundary conditions to standard datasets like Herold et al. (2008), covering: land–sea mask/topography, ocean gateways/bathymetry, prescribed vegetation/land surface, ice-sheet extent/height, CO₂ and other greenhouse gases, aerosols, and orbital parameters. This would enable readers to attribute climate differences (e.g., in heat transport or precipitation) to boundary conditions versus intrinsic model behavior.
Importantly, for any differences in these simulation results compared with previous iCESM1.2 simulations (e.g., temperatures, hydroclimate, or δ¹⁸O), provide mechanistic explanations. Link these to physical processes such as changes in boundary / initial conditions where possible.
2. From a MioMIP perspective, please situate your iCESM1.2 ensemble relative to other MioMIP model results available at https://www.deepmip.org/data-miocene/. Are the large-scale Miocene patterns (e.g., polar amplification, tropical SST gradients, monsoon intensity, zonal mean precipitation shifts, sea-ice extent) broadly consistent, and where does iCESM deviate? Identify explicitly what this study adds beyond existing MioMIP findings—such as new water-isotope diagnostics, refined boundary conditions, a unique climate mechanism, or improved process attribution (e.g., via energy balance, heat transport, or cloud feedbacks).
3. The current Results section reads primarily as descriptive documentation. The manuscript would be much stronger if the authors emphasized the most exciting outcomes—those that either revise our understanding of Miocene climate or demonstrate new model insights. Identify explicitly which findings are novel and discuss their broader implications.
4. Consider merging Section 3.4.1 (“Comparison with preindustrial climate”) and Figure 10 with Section 3.1 and the earlier figures. Since the preindustrial (PI) reference curves already appear in Figure 1, integrating the full PI comparison earlier would make the narrative more cohesive and help readers interpret Miocene–PI anomalies throughout the paper.
5. Including an Atlantic Meridional Overturning Circulation (AMOC) diagnostic would enhance the discussion of oceanic heat transport. A figure showing the maximum Atlantic overturning stream function (and, if possible, a depth–latitude structure) would connect directly to the meridional heat flux analysis in Figure 7. A concise global overturning depiction could also help link the ocean circulation to the modeled heat transport and energy balance frameworks.
Minor comments
1. In Tables 1 and 2, the listed values appear overly precise for the intended context. Rounding to single-digit precision (or justifying the current precision) would improve readability and consistency.