Arctic Ocean sea ice cover during the penultimate glacial and the last interglacial

Coinciding with global warming, Arctic sea ice has rapidly decreased during the last four decades and climate scenarios suggest that sea ice may completely disappear during summer within the next about 50–100 years. Here we produce Arctic sea ice biomarker proxy records for the penultimate glacial (Marine Isotope Stage 6) and the subsequent last interglacial (Marine Isotope Stage 5e). The latter is a time interval when the high latitudes were significantly warmer than today. We document that even under such warmer climate conditions, sea ice existed in the central Arctic Ocean during summer, whereas sea ice was significantly reduced along the Barents Sea continental margin influenced by Atlantic Water inflow. Our proxy reconstruction of the last interglacial sea ice cover is supported by climate simulations, although some proxy data/model inconsistencies still exist. During late Marine Isotope Stage 6, polynya-type conditions occurred off the major ice sheets along the northern Barents and East Siberian continental margins, contradicting a giant Marine Isotope Stage 6 ice shelf that covered the entire Arctic Ocean.

This paper presents Arctic sea ice reconstructions based on semi-quantitative lipid biomarkers from four sediment cores for a time period (late MIS 6 to MIS 5, including MIS 5e) for which such reconstructions have to date not been available. During MIS 5e, or the Eemian, the Arctic was significantly warmer than today, serving as an analogue for the current rapid warming in the High North. The four study sites are well chosen representing different environmental settings, with two sites located in the central Arctic Ocean, the third close to the Laptev Sea continental margin and the fourth site at the Barents Sea continental margin. The study includes highly important findings based on the biomarker data and supported by microfossil data from previous work: 1) During MIS 6, polynya-like conditions most likely occurred off the major circum-Arctic ice sheets, contradicting a previously hypothesised giant MIS 6 ice shelf that would have covered the entire Arctic Ocean. 2) Even during the Eemian, when climate was clearly warmer than today, sea ice existed in the central Arctic Ocean during summer, whereas ice was significantly reduced along the Barents Sea continental margin influenced by Atlantic Water inflow.
The manuscript is generally well-written and clear. I would, however, suggest checking the English carefully once more, as there were mistakes and also somewhat clumsy sentences (some of these corrected in the pdf of the ms).
The main conclusions of the paper are well-supported by the data. This work is definitely of interest to a wide audience in several fields and of an importance and novelty that warrants publication in Nature Communications.
I have added my comments directly to the manuscript. There is one matter in particular I would like the authors to address, namely adding -if at all possible -data on HBI III. I suspect the authors will have analysed this biomarker alongside IP25. Although the nature of this biomarker proxy is not yet fully understood, there are strong indications it could be used as a marginal ice zone (MIZ) / polynya indicator. See for example: 1) Collins et al 2013. Evaluating highly branched isoprenoid (HBI) biomarkers as a novel Antarctic sea-ice proxy in deep ocean glacial age sediments. 2) Belt et al. 2015. Identification of paleo Arctic winter sea ice limits and the marginal ice zone: Optimised biomarker-based reconstructions of late Quaternary Arctic sea ice. The use of such a proxy would certainly further strengthen the interpretations made in the Stein et al. manuscript. Belt et al. 2015 also point out that the use of HBI III instead of a phytoplankton marker (usually brassicasterol) when calculating the PIP25 -index could produce more reliable results as this index does not potentially suffer from the issue of a variable balance factor (more on this in my specific comments added directly onto the manuscript). The work could be reproduced based on the level of detail provided.
Please note also the figures include comments.
Reviewer #2 (Remarks to the Author): Comments on: Arctic Ocean sea ice cover during the penultimate glacial and last interglacial The study by Stein et al explores the environmental conditions of the Arctic Ocean during the last interglacial period (Eemian) and its preceeding glaciation phase (stage 6). Their strategy is to use new data of sea-ice biomarker indices to make assumptions on the existence of sea-ice and openocean ice shelf coverage for these two time intervals. Their motivation probably stems from other studies that claim for a thick ice shelf covering the entire AO basin during stage 6, and an almost sea-ice free summer during the warm Eemian interglacial period. The proxy data are then backed up by some modelling efforts for 3 Eemian time slices for different seasons.
Overall, this is an interesting topic and some of the data are indeed quite unique/new by comparison to previous studies from the Arctic Ocean. Also, the main messages contained in the abstract are timely, thus, the story has potential to be published in Nat Comm eventually.
Having said, the implications of the interpretation of the proxy data remains somewhat shallow, in particular, there is little use being made of the model results leaving the entire story rather underdeveloped. I think most of it owes to the fact that the text strays away, is not really focussed on those 2 critical messages talked about in the abstract. In many respect the manuscript has a strong review character reiterating stratigraphic and other data that have been discussed and published at length elsewhere. It starts with an overly long introduction, but continues like that throughout the rest of the manuscript. The discussional part remains by and large descriptive...

Some further comments:
There are quite a few inconsistencies or loose expressions used; some selected examples to follow.... For instance, line 32, 'sea-cover restricts primary productivity....' Well, there still is plenty of activity under sea ice. Therefore the more precise description should be 'reduces...' Line 51-52: Reference 11 is used to manifest high T in NE Siberia during the Eemian....Well, the study referred to was actually carried out in NW-Central Siberia.... This leads on to LIG sea level rise. Although rather debatable in terms of its timing globally, but it is clear that at some time, certainly still at 130 ka, sea level in the Arctic was likely lower, and not covering the rather wide Siberian shelves. But this has tremendous effects on the spatial size of the sea ice... And, what about the opening of the Bering Strait? I think the modelling part needs to be really improved by introducing 2 more scenarios at 130 ka: one with still closed Bering Strait and nonflooded shelves; another with open Bering Strait and half-flooded shelves....
In the supplementary, I believe some of the data shown are without proper referencing (e.g., Fig.  3); Also, stable isotope data are shown for PS51/038-3, what about the other cores from the central AO (e.g. PS2200) aren't there any such data at all?
In summary, the topic has high potential, but judging the manuscript as is there is a need to have a much tighter focus on the core of the story.

Reviewer #3 (Remarks to the Author):
Stein et al. reconstruct past sea ice conditions from four sediment cores in the Arctic Ocean. They employ the relatively new IP25 method combined with a proxy for ice-free water to ameliorate the ambiguity of IP25 for situations with either permanent sea ice or total lack of ice. Their main results include qualitative reconstructions of sea ice ranging from the penultimate glacial period through the Last Interglacial (LIG) until the early part of the last glaciation (~150 to ~70 thousand years before present).
The results are novel and important. Although the discussion of the results is often speculative and vague, this is in my opinion acceptable here due to the novelty of the data that already represents a significant advance over the previous state-of-the-art. I think this paper is very well suited for Nature Communication and will receive significant attention as the field of sea ice reconstructions develops further and more data is obtained. The discussion of the implications of the new findings could be improved as outlined below. The integration of model results is scarcely motivated and seems like an afterthought. The quality of the figures is good. As a consequence, my criticism is minor and mostly concerns the connection with existing studies. As a caveat, it is important to note that I am not qualified to comment on the robustness or quality of the proxy data.
Given the likely substantial Arctic warming during the LIG (NEEM community members, 2013;Otto-Bliesner et al., 2013), I believe the result that sea ice conditions during MIS5e resembled those of today is quite unexpected. Since the contemporary decrease in sea ice is used repeatedly as a motivation (abstract, introduction), the manuscript should expand the discussion of the implications. If air temperatures really were much higher during MIS5e than today, one of the few remaining ways to maintain a sizeable sea ice cover is a reduced ocean heat transport. Such a change was found in two independent model simulations (Born et al., 2010(Born et al., , 2011. Sea surface temperatures in the Nordic Seas show that ocean heat input from the North Atlantic was higher during the early parts of MIS5d than during MIS5e, where the former has a orbital configuration very similar to today (Bauch et al., 2007(Bauch et al., , 2008Risebrobakken et al., 2005Risebrobakken et al., , 2007. The connection between the original proxy data and the previously published model simulations is not very clear. The purpose of including the model appears to be the confirmation of the proxy results, but since sea ice models are plagued by enormous uncertainties this argument is rather weak. It would be nice to have some more detail on the mechanisms of sea ice reduction in the MIS5e and the RCP scenarios. An energy balance calculation could help to understand why there is more sea ice during MIS5e than in the RCP scenarios and whether this is due to atmospheric or oceanic fluxes. Lastly, the consequences of a reduced sea ice cover should briefly be discussed, also in view of potentially too high terrestrial temperature reconstructions for MIS5e. A suitable recent reference is Merz et al. (2016).
Minor comments: Some laguage editing will be necessary.