The manuscript "Millennial-scale sea surface temperatures of the western Arabian Sea between 37–67 ka BP" by Scott et al. presented excellent new data of paired Mag/Ca SST data from the Arabian Sea. The data covers a critical time in the North Hemisphere that experiences abrupt oscillations of millennial-centennial scales; therefore, it is valuable to test any linkage or temporal or phase relationships between the high and low latitudes climate. The data are well-presented, and interpretations are reasonable and thus publishable in EGUsphere. Though the manuscript is worthy of publication, I suggest the authors show the age model uncertainty of the core (Figure 3) while comparing it with the ice core. Several statistical methods allow the authors to show the C14-based age model with age uncertainties associated with the dating. The revised figures with age uncertainties will be more helpful in examining the temporal relationships between the H events and rapid changes in the Arabian monsoon, and the interpretation could be better incorporated into the text of the next version. The H events should be highlighted in Figure 6, too, and also need to be examined for any temporal relationships between the NH cooling and temperature gradient (upwelling?) in the Arabian Sea. I will be more than happy to evaluate the following revised version of the manuscript and look forward to the publication of the paper. 

Review Scott et al. CLP

The paper re-visits the classical sediment core NIOP 929 in the western Arabian Sea that has been previously used for studying surface ocean and paleoproductivity changes in response to variations in monsoon circulation. The paper extends previous data by adding high resolution Mg/Ca based sea surface temperature records (based on G. bulloides and G. ruber) for marine isotope stage 2 to 4, covering the time-interval of large-scale millennial-scale climate fluctuations as known for example from ice-cores.

The paired Mg/Ca-based SST records allow to distinguish temperatures during both modern winter and summer monsoon seasons with high temporal resolution, clearly sufficient to display the pastern and timing of millennial-scale changes. Both SST records reveal high amplitude fluctuations, being much larger than millennial-scale SST changes during the last deglaciations or the previous interglacial. Importantly, these summer and winter SST changes do not show a systematic relationship with glacial northern hemisphere millennial climate change. This is important as the Arabian Sea and the northern hemisphere monsoon have been traditionally considered being primarily connected with the Greenland millennial-scale pattern. The authors discuss possible scenarios, also involving seasonal shifts of the ITCZ with potential more southern (Antarctic) influence during boreal summer when the convergence zone moves northward.

Overall, the paper is carefully written and the discussions are detailed and nicely involve previous findings form paleo-records in the Arabian Sea. It is hard to objectively assess the timing and pattern of the SST records in terms of northern versus southern millennial-scale timing. I partly agree with the comments of reviewer RC1 regarding a better presentation of the age model. However, radiocarbon dating will probably not provide sufficiently small errors to unequivocally connect to Greenland versus Antarctic pattern as 37-67 ka is at the limit of 14C dating. Therefore, the tuning approach with d18O records is probably the best age control possible for the older part of the record.

As quite some interpretation relies on the visual comparison to the ice-core records, I suggest to include them in all figures 4 to 6 to allow better illustrate the suggested non-straightforward relation of the summer and winter records to northern and southern hemisphere climate changes at millennial time-scales.

Sorry for this long review. I really enjoyed the article, but there are several critical problems. Perhaps some of my comments can't be dealt with, and that's completely OK but please explain why you don't agree wiht my comment.

The article presents a high-resolution, high-quality SST and isotopic data from a marine sediment core collected offshore Somalia. The new data focus on MIS3, and are discussed along previously published data from the same core (but focusing on different time intervals) and other cores collected in the region. The analysis attempts resolving winter and summer SST variability using Mg/Ca measurements performed on planktonic foraminifera with contrasting seasonalities. And that's a difficult task - so I commend the attempt. The authors conclude that the ITCZ latitudinal movements have differential impact on the overall SST signals w.r.t. the hemisphere that controls the signal (northern vs. southern hemisphere in boreal winter vs. summer, respectively) that alter the possibility to read a clear northern vs. southern signal in the seasonal SST regimes reconstructed using both foraminifera species.

I think the article and its interpretation should eventually be published, and that EGUSphere is a perfect journal to publish in. However I think a deep and thorough revision needs to be done before.

First, in a region where such a complex seasonality (summer being colder than winter) without a description of modern climatology in a full chapter is simply not possible. Figure 1 is not helpful in that regard. Please add a full paragraph and figures to describe clearly it.

Second, a deep and thorough description of the age model, with the mandatory uncertainties and an appropriate and fair discussion on the likely very much larger uncertainties for MIS3 has to be added. There not even mention of the s.d. of the radiocarbon dates... In particular in articles dealing with MIS3 the uncertainties of the age w.r.t. the ones applying to ice cores are just so much different that you can't hide this limitation.

Third, the way you present the winter and summer equations have to be fully described in the article. It is not possible to simply refer to other articles. The reader has to have in hands a minimum of background, and certainly with an extra figure showcasing the seasonality in G. ruber and bulloides to evaluate the winter and summer equations without downloading a series of articles.

I really think those three items need to be seriously dealt with prior to any attempt to discuss the science. This being said, I now start the formal review in the order it appears directly in the article.

2.3. (Age model): I don’t think the authors can justify a fine-tuning of d18O to NGRIP as long as they conclude that SST don’t look to any of the ice cores. Please discuss a bit more this processing. Also, please show the original d18O values and the result of the filtering on the same figure and Y axis. It is unclear which dataset is already processed.

2.5. (W and S SST): I find the r and b values strange and arbitrary in Saher. Please discuss this issue here. In general, I don’t like the way to assign some weight in the W equation with both species and the W into the S SST estimation equation. It doesn’t make any mathematical sense to me. Playing around with this set of equations could drive the math towards uncomfortable solutions. For example, very small changes in the numbers of the parameters lead me to estimate S SST uniquely with ruber... but it is mean-annual. Also, the S SST as it stands depends partly on the W SST, which doesn’t make any sense to me either, etc. So, please, even if I won’t agree with your interpretation in the end (I’m fine with that, and will accept your way of interpreting), at least please elaborate more discussion on the choice of your parameters because the resulting W and S SST depend crucially on the exact value of these parameters.

3.1. (d18O): in fact, I’m not sure that I understand it well: are the d18O are smoothed or not? Please show first the raw values.

4.1. (robustness…): please also make sure in the text and interpretation of different species and Uk’37 to be as firm as possible. You sometime state that Mg/Ca G. ruber and Uk’37 represent mean-annual SST. Then you may also elaborate why this might not be valid anymore during the MIS5, that by the way implies the season you assign to different proxies are not constant over time, that adds degrees of freedom in the interpretation to the seasonal SST changes and associated mechanisms described.

Line 210: if you have the SST estimation, a very simple calculation given the calibration equation used in Rostek will give the Uk’37 you want to calculate to investigate different calibrations. So yes, that is a very basic option.

Sentence lines 216-218: I don’t understand your point here.

Sentence lines 222-224: Again, a specific paragraph is absolutely needed to describe the different foram fluxes you cite here.

Line 230: it is curious: there is no foram weighting prior to foram dissolution before analyzing Mg/Ca? This should have been a very basic/classic data reported somewhere.

Sentence lines 232-234: I don’t agree with this statement. In this region, alkenones often/always provide a point-to-point scattering that is much lower than Mg/Ca for many reasons even when you incrase resolution.

Discussions lines 244-262: please try to structure the statements, it goes a bit in any direction. For example, again, if G. ruber tolerates everything, stil its Mg/Ca fluctuates. The fact that it is mean-annual has repercussions on the meaning of your seasonal index I criticized earlier. I also feel a larger discussion of the Ganssen data (on individual foraminifera analysis from the same species) should be detailed very much more, there is some important take-home message for you in their analysis on seasonality.

4.2. (long-term…): dealing with H4, well… please be clearer on the limits of your age model. Clearly, the H4 grey bar on the NGRIP could be larger and finish right on the following D/O event, while the corresponding grey bar width on the d18O of ruber is… unknown. Specifically for H4 it has been already described that the H4 has two distinctive events, so the grey bar width could perhaps be expanded very much, having quite an implication on your discussion on the seasonality you describe for your site. Then, dealing with the SW/NE monsoon evolution, interesting model simulations available for the Holocene in Bassinot et al., 2011, Climate of the Past, could help describing the shifts in foramfluxes for different species under varying climate conditions. 

Sentence lines 289-290: again, you may cite Ganssen and discuss his data. Also, the fact that Uk’37is even warmer than G. ruber might reveal in fact that Uk’37 is NOT the mean-annual SST.

4.3. (phase of…): again in the Naughton (EPSL, 2009) article there is clear ‘double H4’ signature. It may allow you enlarging your sedimentary sequence thickness of your own H4 and changes your interpretation.

The ITCZ mechanism described is interesting, but already described in length in the western tropical Indian Ocean using G. ruber and Uk’37, model simulations etc. (see Wang et al., 2013, Paleoceanography). You really have to cite that article of you keep the discussion as it is (that I liked).