Summary

The study by van der Weijst et al presents multiproxy-based temperature and productivity records spanning the past 15 Ma. Data are generated from an ODP core recovered from the eastern equatorial Atlantic, at a site that is under the influence of monsoon-induced upwelling. The proxies used are alkenone-based UK’37, archaeal GDGT-based TEX86 and dinocysts. Based on several lines of evidence, the authors argue that the TEX86 proxy records a mixture of surface and subsurface signal, likely that of Antarctic Intermediate Water (AAIW) as also suggested by two previous studies. Assuming that UK’37 reflects SST variability, benthic d18O records NADW variability and TEX86 reflects AAIW variability, in combination with other published temperature records and a pCO₂ record, the authors discuss possible mechanisms that lead to M2 glaciation.

General comments

Overall I find the manuscript very clear and well-written. The topic also fits well within the remit of the journal, thus will likely be of interest to the broad readership of Climate of the Past. Although generally accessible to the reader, I think some arguments can be further improved and/or need further clarification. I hope the authors will find my comments and suggestions helpful in revising the manuscript. Altogether it should amount to moderate revision. Below I outline my major concerns.

(1) TEX86 reflects AAIW variability

The authors provided several lines of evidence to support their claim that TEX86 reflects subsurface temperature: high GDGT 2/3 ratio, TEX86H SSTs that are out of the modern range and unrealistically large magnitude of change over the last 15 Ma, temporal trends that are more similar to those of benthic d18O than UK’37 and Mg/Ca based on both mixed layer- and thermocline-dwelling species. The discussion is generally convincing.

What is however unclear to me is how the authors then link the TEX86 signal to AAIW variability. I find the explanation provided at Line 226-227 to be rather vague. The core-top value of their SubT TEX86 record is ~14ºC, which is a lot warmer than the temperature in AAIW (~5ºC according to Figure 7a) and is in fact closer to that of South Atlantic Central Water. Further, the core of depth range integrated by the HL16 calibration is 100-350 m, which is much shallower than the average depth of AAIW. As the climate interpretation hinges on this claim, the authors need to provide clearer arguments to support their statement. I wonder if it is possible to attribute the relative influence of SACW vs. AAIW using the depth distribution in the calibration used?

(2) Subsurface export of GDGTs

Most sedimentary GDGT 2/3 ratios along the core are > 5. The authors argue that this indicates that GDGTs in the sediment core are partially sourced from the subsurface ocean. I acknowledge that the GDGT 2/3 ratio is a routinely used indicator to cull TEX86 data, it does however worry me that the correlation between this ratio and TEX86 is quite strong for this core. The authors use the positive correlation to argue that the TEX86 variability is NOT controlled by subsurface-sourced GDGTs, as more subsurface GDGTs should result in colder temperature estimates hence also a negative correlation between TEX86 and GDGT 2/3 ratio. But alternatively this positive correlation might simply mean that the ratio is indeed reflecting temperature change instead of changes in the surface vs subsurface source of GDGT. Indeed, from Figure 7 of Taylor et al. (2013, GPC), it does look like the GDGT 2/3 ratio is not only correlated to water depth but also to SST. If true, this would mean that the authors lose one of the strongest lines of evidence for the subsurface origin of their TEX86 record. I encourage the authors to discuss this possibility in detail.

Second, what process would be responsible for the export of GDGTs from subsurface ocean at 350 m or within the AAIW to the seafloor? I appreciate the fact that this is not a calibration study, but the lack of vehicle to export subsurface ocean GDGTs to depths is one of the main criticisms for the HL16 calibration, so I think the reader will be a lot more convinced if the authors can propose some possible mechanisms to explain/demonstrate that a subsurface export of GDGTs is indeed possible.

(3) Is the temporal resolution of the records sufficiently high to assess a 5-10 ky lead/lag relationship between records?  

The paleoclimate interpretation is largely based on the temporal patterns, so the choice of calibration does not matter much. But what matters here instead is the temporal resolution. Most of the discussion in the last sub-section of Discussion is based on the 5-10 ky lead of TEX86 over benthic d18O, and the lead/lag relationship between pCO2 and temperature records. The lead of TEX86 over benthic d18O is, to my eyes, based on one or two data points. Having said that, I don’t think that one can make such a claim given the low-resolution of the proxy records and uncertainty in proxy measurements. I urge the authors to give this some more thought and provide a more balanced discussion taking into account the limitations of their dataset.

Specific comments

Line 13-15: I would flip the arguments around. First the ecological evidence that Thaumarchaeota / GDGTs occur in the subsurface ocean, then only the circumstantial evidence of the good correlation of TEX86 to temperature from various depths.

Line 17: “proved” is a bit strong. In my opinion, structural similarities in downcore records are at best circumstantial evidence, not direct proof. What about “can be best assessed in downcore studies”?

Line 166-174: The authors argue that TEX86 records subsurface ocean temperature variability because the long-term trend in TEX86 differs from that of other SST proxies, including UK’37 and Mg/Ca. As the authors noted, these two proxies have their own issues: UK’37 is close to saturation and thus may be insensitive to temperature change, whereas Mg/Ca is susceptible to secular change in seawater Mg/Ca, carbonate chemistry and salinity. I note that the Mg/Ca records are from a sister paper (cp-2021-68), and likely the authors have discussed all the issues in that paper. It would still make life easier for the reader if the authors can briefly summarize these Mg/Ca issues here and why the trend is robust despite these potential caveats. As for UK’37, it might be helpful to also test the Bayspline or other tropical UK’37 calibration (e.g. Sonzogni et al. 1997 DSR II) that has a steeper slope than the one in Prahl and Wakeham (1987 Nature) or Müller et al. (1998 GCA).

Line 189-190: Unclear reasoning, please rephrase. Also see my general comment (2).

Line 206-207: It is difficult to see this in Figure 5. It might be helpful if the authors can illustrate the lead in the figure. Also see my general comment (3).

Line 223–224: Again, this is not immediately clear from the figure. Also, how is it established? How is the onset defined? See comment on Line 206-207.

Line 240-241: I would rephrase this. Strictly speaking, HL16 calibration does not assume/target any water depths. Instead, they search for calibrations that can reconcile the variability of UK’37 and TEX86. The depth distribution indicates that the most probable depth range is 100-350 m. Thus, instead of using the entire calibration ensemble that includes all depth ranges down to 950 m, one may very well choose one that is calibrated to 100-350 m.

Line 242-244: Thaumarchaeotal cell counts and GDGT concentrations vary a lot in space. Are studies from the South Atlantic and Arabian Sea the suitable choice of reference here? If no water column studies are available in the study area, at least tell the reader why it is reasonable to assume that what happens in the water column elsewhere might be applicable to eastern equatorial Atlantic.  

Line 251-257: Given the hydrography at the study site which is under the influence of episodic monsoon-induced upwelling, I would expect the temperature variability here to be larger than that at pelagic sites like the warm pool. So I think it unlikely that the 1:1 relationship between the surface and subsurface ocean would hold over millions of years, nor is it reasonable to assume that the subsurface temperature variability here would be comparable to global SST change in the tropics.

Line 277-285: See general comment (3).

Review of

A fifteen-million-year surface- and subsurface-integrated TEX86  temperature record from the eastern equatorial Atlantic by Carolien M. H. van der Weijst et al.

Summary:

The manuscript presents a new 15 Myr TEX86 record from ODP Site 959 in the Gulf of Guinea. Comparing the TEX86 record to other proxies of the same core and other sediment cores, it investigates the source of the recorded TEX86 variability and concludes that for this site, TEX86 is mainly a subsurface temperature proxy. Applying a subsurface calibration, it then discusses the climatic implications of the record and suggests that the M2 glacial was marked by AAIW cooling during an austral summer insolation minimum, and that decreasing CO 2 levels were a feedback, not the initiator, of glacial expansion.

The manuscript is well written and fits in the scope of Climate of the past. The paper is twofold; Based on the multiproxy comparison / interpretation, it provides an attribution of the source for the Tex86 signal for the specific core and supports a rather controversial hypothesis; that the Tex86 signal could originate from subsurface temperature variations and would thus also have a different temperature sensitivity than assumed in most studies. In addition to this proxy attribution part, it also provides a paleoclimatic interpretation of the record which is largely independent on the calibration but hinges on the timing of the records.

I have only three comments and would recommend the manuscript for publication in CP

• The manuscript concludes that the Tex86 record for this site should be interpreted as subsurface temperature (SubST) and provides a range of convincing arguments supporting this interpretation (magnitude of variations, similarity to benthic records, magnitude of trend, high GDGT 2/3 ratio). Despite this, in large parts, the manuscript clings on the SST interpretation and discusses Tex86 SST and shows Tex86 SST (Line 137, Line 164, Figure 2,3,4,5). At least, to me this is confusing and inconsistent as either the authors interpret the record as subsurface temperature and the SST interpretation is only used as a first guess before the depth attribution; or the authors interpret the record as SST… which would than change large parts of the manuscript and the conclusions. I would thus suggest clarifying that the first result is the Tex86H record (without calibration) and the calibration itself is an interpretation. As an example, in Line 135, instead of ‘Between 15 and 11 Ma, TEX86H-SST  fluctuates…”, one could write “Interpreting TEX86H as SST proxy, the inferred SSTs (here called TEX86H-SST) fluctuate. For the figures, the authors should check if the SST interpretation is needed in all of the figures.

• In several occasions, (e.g. L25, L264, L315) the authors argue that the “depth-integrated TEX86 record can potentially be used to infer SST variability, because subsurface temperature variability is generally tightly linked to SST variability”. However, they also find that the time-series of the Tex86 record differs from the surface/thermocline temperature derived from other proxies and argue that this is due to the depth-integration of their record. Both statements seem contradictory. As support, the authors (L249) cite Ho and Laepple, 2016: “The ratio between temperature change in the surface and subsurface ocean is 1:1 when averaged across many sites and on longer timescales”. However, as the authors show themselves in their presented multiproxy records, this does not apply to any single site or on details as the phasing of temperature variations. It is also unclear why such a translation to SST would be needed as a subsurface proxy also provides important information on the climate dynamics; can be used to validate models etc. Thus, I would suggest removing the parts on the possible translation of SubST to SST or to provide arguments why the authors think such a conversion is possible and useful as the conversion is clearly not just an offset.

• One main argument for the choice of the calibration is the amplitude of the reconstructed variations. This is mainly demonstrated in Figure S1. As the choice of the calibration is a major result of the manuscript, I would suggest moving this Figure into the main manuscript.