The authors present a compelling paper discussing the disagreement between estimates of Miocene to Early Pleistocene upper ocean temperatures in the South Atlantic. Not only do they identify the proxy-proxy disagreement, they identify several sources of biases which may be responsible for said mismatch. In my attached notes, I mention several areas where I believe the text could be improved by providing some methodological clarifications and emphasizing the importance of this work for determining latitudinal temperature gradients, and the broader impacts that those gradient estimates have for our understanding of past climates.

Stoll et al. present an interesting multi-proxy comparison of upper ocean temperatures over the last 15 million years, combining new coccolith clumped isotope-based temperature estimates with published alkenone-based estimates from the same sites. The fact that both proxies are produced by the same organism, but otherwise based on very different principles, makes this comparison especially interesting because it rules out several possible reasons for discrepancies. Yet, the paper presents a huge difference in temperature estimates from the two proxies. Several lines of reasoning including comparison with other evidence suggest that the colder clumped isotope based estimates are more realistic (though possibly slightly cold biased), whereas the Uk’37 based estimates probably suffer from a substantial warm bias. This result has important implications for the interpretation of previous alkenone-based temperature estimates from high latitudes and thus for previous estimates of latitudinal temperature gradients.

The data and message presented here are very clear and I have only minor suggestions for further improvements.

Section 4.1.5 about a suggested warm bias of the Uk’37 data from Site 1088 could be made clearer. If I understand correctly, the bias is suggested to only be present at Site 1088, not 1090, because at the latter site a different index is used (Uk37). This interpretation could be made clearer in this section and it would help non-alkenone-experts like me if the difference between different alkenone indices was introduced earlier in the paper (e.g. before the data are presented in Figure 3). I was also left wondering whether it would be possible to recalculate the respective other index out of raw alkenone data from either site. Another question I have is whether the potential calibration bias due to calibration versus SST rather than production depth temperature (mentioned in section 4.1.1) is specific for either of the two indices or applies to both? It would also be good to know whether using different alkenone calibrations (e.g., culture-based) change the picture in any significant way? In the same vein, it could be added that on the clumped isotope side, any other calibration choice would make the D47 temperatures even colder.

Section 4.2: The comparison with the proxy difference at Sites 1171 and 594 seems to be better placed earlier as additional argument for warm bias in the Uk’37 estimates, rather than in this section about latitudinal gradients. That would allow this section to focus on the implication for reconstructed latitudinal gradients, introduced in the introduction as a major motivation for this work. The figure (Fig 6b) could instead be incorporated into Figure 3.

Minor suggestions and typos

Line 22: «must» – maybe change to «should”, if it could in principle be possible that calcification occurs at somewhat different water depth than alkenone synthesis

Line 45: Check TEX86 spelling

Line 46: Check reference formatting

Line 51: Add a comma after “thermometry”

Line 92-93: Check reference formatting

Line 95: analyzed

Line 101-102: Check sentence structure

Line 111: Check reference formatting

Line 114: Is it correct that each sample replicate was corrected with the closest 12 standard measurements (ETH 1-3), i.e., less that a run worth of standards?

Line 116: What is meant by “batches”?

Line 124: Give references for the alkenone records

Line 141-142: How are the samples containing detrital carbonate treated for interpretation? Are they disregarded (as the crosses in Fig 2 suggest)?

Line 148: At ODP Site 1090,…

Line 191-192: Can it be ruled out that coccolith and alkenone production occur at different water depths? Is this statement linked to the second sentence of this section, possibly implying continuous production of both phases during the lifetime of the organisms? If so, that could be made clearer.

Line 193-195: Specify that the calcification temperature is known for the D47 calibration as it is based on cultures.

Line 196: Add space after “depth”

Line 197: Add space after “temperatures”

Line 199-202: relating the alkenone index to SST or even summer SST even though the signal is produced deeper in the water column or in a different season must assume that vertical or temporal differences are similar at all sites used in the calibration and where the calibration is applied, which seems problematic and could be made even more clear here. Even though the gradient is weak at Site 1088, it could be stronger in most core top locations from the calibration, which would still bias the signal too warm.

Line 204-207: Is this only true/known for part of the record?

Line 216: Check reference formatting

Line 216-217: Could there be other particles of similar size as the coccoliths that are produced at colder temperatures/at the seafloor?

Line 217: Add “in our record” or “here” to make it clear that the results from this study are referred to.

Line 231: Add space after “biomineralization”

Line 266: Diagenetic processes? Instead of “cool … signals” I suggest reformulating to “introduce a cold bias to…” (same for the next part of the sentence)

Line 268: Check formatting of Uk37

Line 271: Add space after “index”

Line 288-289: Check sentence structure

Line 311: Check sentence structure

Line 314: Check sentence structure

Line 316: Check sentence structure. Here and subsequent sentences: Check use of U^k₃₇ versus U^k’₃₇.

Line 320: The reference to a potential salinity effect comes a bit out of the blue and makes the sentence convoluted. Rather split up the sentences and introduce this potential effect better.

Line 326: Check sentence structure with double brackets

Line 335-336: the scale of suggested warm bias would be better established in the previous section where the basis for this statement is discussed, and then repeated here.

Line 347: Check sentence structure

Line 350: If the Uk’37 index is saturated at Site 926, the calculated gradient should be a minimum estimate

Figure 2:

The figure caption could give some more detail on the different kinds of data shown in the figure to allow the reader a quicker overview. E.g., that the species listed here are dominating the respective size fractions (with reference to the supplement), that crossed out samples contain abundant detrital carbonate,…

Caption for Fig 2: Panel b) is for Site 1090, not 1088.

Format axes titles consistently for all Figures (e.g. “Age (Ma)”, “D47 temp. (°C)”)

Figure 4: Figure caption: Check sentence for panel a).