Review comments on “A transient CGCM simulation of the past 3 million years” by K.-S. Yun et al.

The manuscript presents an unprecedented effort to simulate the Pleistocene until present and beyond. It is the first quasi-continuous simulation using a comprehensive coupled general circulation model over a time span as long as three million years. This approach allows the authors to fill a gap between ultra-long simulations using models of intermediate complexity (EMIC) and time slices or short transient runs with fully coupled atmosphere -ocean or Earth system models. The CGCM CESM1.2 used in this study is well suited. Although of coarse resolution, it has been used since many years and has been shown to give satisfying representation of past and present climate states and variability (e.g. ENSO).

The technical design of the experiment is impressive and well thought-through. The authors make use of a well-established acceleration technique I combination with running chunks of the simulation in parallel.

The authors show that the model reproduces the climate evolution in accordance with reconstructions and earlier EMIC simulations (from which, however, the forcing driving the CGCM was derived). More importantly, the authors do a very good job highlighting the added value of the CGCM simulation: the ability to look how atmospheric dynamics and climate variability change under orbital and greenhouse gas forcing. In particular, I appreciate the investigation on the meridional heat transports in ocean and atmosphere. Here the authors come up with very interesting new results.

On a more critical note, the authors could discuss in more detail shortcomings and limits of their experiment. For example, the static LGM topography and land-sea mask. Another important limitation is that there is no coupling of ice-sheet changes and fresh-water forcing of the ocean. This is mentioned in the AMOC section, but it would be more important to discuss the constraints on the meridional heat transports in the ocean under AMOC breakdown and recovery.

Overall, the manuscript is mature and written in a concise and clear manner. The figures are of good quality and convey the message well. I have only identified a very few minor issues (see below).

I therefore recommend accepting the paper with (very) minor revision.

Minor issues:

Line 175ff: could you comment on the discrepancy between “synthetic Greenland” and the simulation, i.e. the much larger amplitude of the reconstruction?

Line 214: the “intensification in amplitude after the MPT” in the East Asian monsoon: this looks more like a gradual increase from 1500 kaBP. Could you be a bit more qauntiative here, e.g. by showing running standard devitations?

Line 294: AMOC and ocean heat transports (see above)

Figure 2: caption more precise: relative to the mean over the entire 3 Ma.

Figure 9: caption more precise: standard deviation over sliding time window

Figure 10: include information how the characteristic modulation (labels in fig. 10 c) have been derived

This paper describes the first transient CGCM simulation of the last 3 Myrs, with a forcing acceleration 𝑓actor of 5. The simulation is compared with proxy observations in various climate regimes and in various climate variables.  This is a tremendous effort and believe the simulation will be a great resource for the community in future studies. The paper is well written and should be published.

I have several minor comments.

• To better quantify the bias caused by the 5x acceleration, it will be good to have a comparison with a transient simulation without acceleration for one time period (say, about a precessional cycle length of about 20kyr).

• Clarify if the temperature discussed in the text and shown in the figures are all annual mean.

• For the model data comparison of long time series, such as those in Figs.4-7, it will be good to have a parallel figure (say, to the right of the time series panel),which show two spectra, one before and one after the MPT. As it stands, it is becoming very hard to judge.