This is an interesting premise for a paper. But ultimately it didn’t deliver.

The premise on the ice sheet melting is not really dealt with - I was expecting the authors to get back to it at the end and show how their new records might link to it. Furthermore it is unclear to me why they focussed on the last 800 ka - they didn't really try to interpret anything older than 400 ka for the interglacials. Why only 7 records - there are others in the literature for LCDW already published that go back to ~ 800 ka e.g. ODP 1168, ODP 1170, ODP 1171 from the South Tasman Rise, south of Tasmania (Nurnberg et al., 2004). I would suggest the authors look at the new dataset of d18O recently published by Mulitza et al., 2022 ESSD, which would have many more records covering the last 400 and even 800 ka. Given the lack of records that go back 800 ka why not focus on the shorter time periods – especially when there are no periods warmer than present between MIS11 and MIS19. They compare to several other datasets that do not cover the last 800 ka in their discussion.

Additionally, the new compilation shows strong agreement to past DWT compilations - so I'm not entirely sure that is really adds much to the literature. I was really hoping it would get into how it might influence the ice sheets or modelling so that it did add something new.

I also find the inclusion of the NADW records (4 out of the total of 7) a little odd given the focus on the LCDW and evidence in the literature of the shut down of AMOC and therefore I would not expect the NADW to be a strong contributor to LCDW during the glacials. The authors came to this eventually in the paper, but only after putting it all together and didn’t really discuss the implications on their records which have 4 records from NADW that may well swamp the LCDW datasets.  Although I realise the authors are interested in the interglacials for the ice sheet melting. But then didn’t really come back to this and how the MIS5, 9 and 11 may have impacted the ice sheets.

Many sections there was insufficient detail on the methods and the issues - they came back to many of these in the discussion at the end - but they should have been upfront. There really was insufficient detail around the methods used for the Mg/Ca and the d18O. I realise these are in the original papers – but you need to provide a summary here so the reader doesn’t have to go back to the original papers.

I felt the oceanography background needed more information and detail on the CDW – and differentiate between the NADW, LCDW, UCDW and mCDW – it would have helped to have a 3D/Depth figure to show the links between these water masses.

It was unclear to me how the authors plotted up Figure 4 when you can’t look at BWT on both the NADW and the LCDW at the same time? How do you make plots like this? Also which cores had both benthic d18O and the Mg/Ca done on them to make the other X-Y plot.

The reason why there are not many DWT records using Mg/Ca for benthic foraminifera is that it is not a trivial thing to do. Firstly there are not always sufficient tests of the right species of benthic foraminifera. Secondly the method is time consuming and has some issues – some of which were outlined in the limitations. But one thing that has not been mentioned in the paper is the potential impact of dissolution and the Carbonate Compensation Depth - the LCDW of the Southern sits at or very close to the Carbonate Saturation Horizon which is around 3000 m and the CCD which sits around 4000 m- so not many cores actually preserve sufficient Carbonate organisms or the records can be compromised by period of dissolution due to the shoaling of the CCD - during the glacial cycle.

Chandler and Langebroek compiled 7 deep water records (3 benthic Mg/Ca, 1 ostracod Mg/Ca, and 3 benthic d18O) during the last 800 ka, aiming to understand the glacial-interglacial temperature variability of Circumpolar Deep Water. However, there are several problems with data selection, data interpretation, and presentation, I have to say that this study, at least in the current form, is not to the standard of the CP.

Here are my main concerns.

1) The premise of using a compilation of BWT at selected sites to reflect CDW BWT change.

Including NADW sites in the compilation is not appropriate because 1) NADW sites could be bathed in different water masses during glacials and interglacials due to circulation change, as already mentioned by the authors in Section 5.3, 2) shallow NADW sites (e.g. ODP 980) would have minimal influence on CDW that can potentially upwell underneath the Antarctic Ice shelves, due to the lower seawater density. On the contrary, Pacific sites at relatively deep depths, though downstream of CDW, may more reliably record the CDW temperature as those could be on the same isopycnal as CDW upwelling underneath the Antarctic Ice shelves. More benthic d18O records from the Pacific included in Bates et al (2014) thus might be included in the compilation as well. But ultimately, it is not most convincing to include sites not bathed in CDW to infer CDW changes. There may be more SO sites available if the time span of the compilation can be shortened.

2) Using a compiled CDW T to inform ice shelf melting triggered by CDW.

The authors set out to use a compiled CDW T record to infer potential ice shelf melting triggered by CDW during previous warm interglacials. It is noted that such an event can be triggered by ~1 degC warming in a short period (a year) in the modern ocean (e.g., Jenkins et al., 2018), which is a really small difference challenging for any given record reconstructed by any proxy to resolve with confidence, let alone a compilation smoothing multiple records. So the 0.1-0.5degC warming during previous interglacials mentioned in the abstract must be interpreted in the context of the uncertainty range, which is not reported but can be expected to be large enough to make the reported warming statistically insignificant.

3) Hydrographic settings of CDW.

As a manuscript on CDW, there are lots of inaccurate statements about CDW. While NADW is the heat source of the modified CDW, it is not the principal source water mass of CDW (Line 7). Deep waters from the Indian and Pacific sectors, as well as deep water formed around Antarctica, are also key sources of CDW (refer to Talley 2011 for detailed hydrography). Also, UCDW and LCDW are not defined properly. And modified CDW is often used in the context of melting under ice shelves, but it is not a counterpart of UCDW and LCDW.

4) Lack of details about their method.

how the BWT is calculated? How is the sampling done? Which types of uncertainties are included in the uncertainty envelope shown in Figure 4? These questions are not trivial. Details, such as choice of d18O-T equation, treatments of ice volume change, and local effects on seawater d18O, are lacking for benthic d18O calculation. For the sampling method, no original data is shown, so one would have no way to assess if the compilation is potentially biased by one particular site.

5) The structure of the paper.

A reader is left with the impression that the work done here did not at all contribute to solving the research question raised in the introduction, i.e., the potential influence of CDW T on ice shelf melting during previous interglacials. The introduction sets the expectation of a reader very high and by finishing reading I felt a little disappointed. Perhaps, the authors could put something that can be achieved by this work in the introduction. Also, I think it would be better to provide some background on the limitations of the employed proxies in the introduction.