Comment 1: Section 4.1. The authors describe the biostratigraphic events that they found in the studied samples, however, I do not agree with the way the results are reported. For example, the last occurrence (LO) of altispira was located at sample U1575A-5R-7W-0-2 cm (46.89 mbsf) but since the authors didn’t study any sample between 46.89 and 44.18 mbsf they should give a depth range for the LO of that species. I mean, the LO should be placed between 46.89 and 44.18 mbsf, in the mid-point with an error bar covering the whole interval. Same thing for the rest of the events. The use of “top occurrence” is not very common, I suggest using LO or LAD. Also, please revise the definition of the biostratigraphic zonation, for example the LO of D. altispira does not indicate PL3-PL4, but indicates PL4-PL5.

Author response: we thank the reviewer of these observations. We believe that the use of the mid-point depth for a certain bioevent means that the uncertainty of the next bioevent already could potentially merge with it. In turn, this indicates that an age-depth model constructed with this approach becomes more unreliable. Moreover, it is always important to record where the lowest and highest occurrence of a species within the core actually occur. For the above-mentioned reason, we prefer not to consider the mid-point depth solution. However, we agree with the reviewer that the errors in the placement of the biostratigraphic datapoints must be taken into account. Thus, we will add in the two tables an error value for each biostratigraphic event, to express the uncertainty in the placement of the datums. We believe that the use of the terms “base” and “top” for the occurrences of taxa within a specific stratigraphic sequence should be preferred. In fact, the acronyms FAD and LAD must be avoided as they indicate a global event or the occurrence of a taxon in a biogeographic region. However, speciation “cannot be considered an instantaneous event and time difference between the evolution of a new species in some given location and the time in which its distribution reaches its widest extent, may or may not be global” (see Smith et al., 2014). Thus, it is more appropriate to refer to local occurrences. In this respect, the use of the first occurrence (FO) and last occurrence (LO) could still induce misunderstandings in the interpretation of the bioevents as local. Moreover, Base (B) and Top (T) are commonly used for both planktonic foraminifera and calcareous nannofossils bioevents (see Lam and Leckie, 2020; Wade et al., 2011; Backman et al., 2012; Agnini et al., 2019). The LO of D. altispira define the top of the PL4 zone in both Gradstein et al., 2020 and Wade et al., 2011. The assignment of the sample to planktonic foraminifera zones PL3-PL4 is based on the concomitant presence of Dentoglobigerina altispira and Globoconella inflata (see lines 398-399).     

Comment 2: Figures 3 and 5 could be improved if the paleoceanographic information is included (e.g. cluster 2= normal BOC conditions) either in the caption or in the figure as a legend. I think the figures would be clearer if the same colors are used for the clusters in both sites (e.g. cluster 2 always in green). And the same color can also be used for figure 6. Also in the text, it would be clearer if in both sites the same nomenclature is used for similar paleoceanographic conditions (e.g. in U1575 Agulhas leakage is cluster 3 and in U1576 it is cluster 1b, it would be better to use the same number). Numerical age based on the biostratigraphic zonation would be helpful to better understand the paleoceanographic evolution.

Author response: We kindly thank the reviewer for these suggestions. We will add the paleoceanographic information associated with all the clusters in the figure caption or as legend in the graph. The use of the same colors for the clusters at both sites would be misleading as they do not exactly indicate the same paleoceanographic conditions. This is shown by PCA results and by the relative abundance data of the planktonic foraminiferal assemblages. Specifically, relative abundances of the main species defining the clusters are similar but not completely equal and this discrepancy could be explained by longitudinal and latitudinal variations in the thermocline fluctuations and water mass interactions. Those subtle differences were detected because of the key locations of the investigated sites, with Hole U1576A situated in a more southern position and closer to the center of the gyre, compared to Hole U1575A. Regarding the cluster nomenclature, the different letters and numbers for the two sites depended on the clustering procedure and, thus, on the placement of a specific cluster in the dendrograms. We believe that, adding the paleoceanographic conditions associated with each cluster in the figure captions/legend (as recommended by the reviewer) will make the comprehension of the paleoceanographic evolution sufficiently clear for the readers. Age intervals will also be provided in Figure 3 and 5.

Comment 3: Please include a sketch for cluster 3 (Agulhas leakage) in figure 6.

Author response: We are grateful to the reviewer for this observation. We will add a sketch of the Agulhas leakage condition in Figure 6. 

Comment 4: Could you correlate the conditions identified with the clusters in both sites and make a more comprehensive paleoceanographic reconstruction? How does this dataset correlate with the previous study in the region by Ufkes and Kroon (2012)?

Author response: We are unsure what the reviewer means by correlating the paleoceanographic conditions in both sites. If the reviewer refers to a time correlation, unfortunately this cannot be feasible due to the low biostratigraphic resolution. We cannot be sure that, within the same biostratigraphic interval, the paleoceanographic conditions at Site U1575 and U1576 occur exactly at the same time. Regarding the correlation between our dataset and the results from Ufkes and Kroon (2012), the different biostratigraphic resolution of the two studies does not allow to perform a detailed correlation of the paleoceanographic conditions over time. Furthermore, Ufkes and Kroon (2012) focused their study on the last 1.1 Ma, not covering the whole EMPT (as specified at lines 45 and 46). However, we recognized similar paleoceanographic settings (represented by the different clusters) such as an intensification of the upwelling phases (see lines 631-634) and the shift of the ABF front, with the latter detected based on the change in abundances of G. crassaformis linked to the Benguela Niño events in the Southeast Atlantic Ocean (please refer to lines 672-679). 

Minor suggestions from reviewer 1

Comment 5: Line 287, Discoaster must be in italics

Author response: We thank the reviewer for the comment and we will apply the requested change.

Comment 6: Line 291, The total number of foraminifers does not give us a lot of information. I suggest to change this sentence by the average number of specimens counted per sample, indicating also the minimum and maximum number of specimens counted per sample.

Author response: we thank the reviewer for the useful comment. We will change the sentence “A total of 6996 and 5540 planktonic foraminiferal specimens were identified for the assemblage study in Holes U1575A and U1576A, respectively.” with “An average of 304 (min=283; max=327) and 308 (min=297; max=346) specimens per sample were counted and identified for the assemblage study in Holes U1575A and U1576A, respectively”.

Comment 7: Line 672, change Hole by Sites

Author response: We thank the reviewer for the comment. We cannot substitute Hole with Site as according to the IODP terminology the expression “Holes U1575A and U1576A is the correct option. However, we noted that there is a plural form missing at line 672. Thus, we will change the term “Hole” with “Holes”.     

Comment 8: Table 1, I think it is better to define the biostratigraphic events as LO (last occurrence) and FO (first occurrence instead of Top and Bottom, although I understand that you are using the same terms as in Wade et al. (2011).

Author response: We thank the reviewer for the suggestion. Please refer to the previous reply for comment 1. 

Reviewer´s comments:  Before writing my comment to the authors, I would like to emphasise that I am a geochemist, paleoceanographer and climatologist first and foremost. I would therefore urge caution in interpreting the following review.

My reading of the manuscript suggests that the authors are well-informed taxonomists, and some of them were involved in the IODP Expedition 391. However, I am not a good person to judge the quality of the dataset presented in this study, as I am a poor taxonomist. Reading Reviewer #1's comment suggests to me that the data are well handled, nicely analysed, and I have no doubt that the article could be published somewhere.

However, as a frequent reader - and reviewer - of Climate of the Past, I do not think the article is appropriate for that journal. I suggest the authors submit it to a more appropriate journal to publish the data set, methodology and interpretation.

As it stands, it was almost impossible for me to rely on any "big scientific questions" in the field of palaeoclimatology. The article is very descriptive, is not organised to convey any important advances in paleoceanography, and reads like a preliminary report from the IODP expeditions. The article clearly lacks:

- a readable time frame

- a clear unanswered scientific question in the broader field of palaeoclimatology

- a comparison of the results with other results published in the literature to answer the scientific question (the data aren't even presented on a time scale, so we never really know what time window we're looking at!): you cite a lot of googled SST records, but never compare your results with key results already published in the field on an appropriate figure.

In fact, the introduction and conclusion list a number of objectives, but fail to resolve any of the burning questions that the field of palaeoclimatology has been chasing for decades. Figure 6 is certainly interesting, but as it stands it can't really be useful unless the reader has some idea of the time frame we're looking at. I ended up putting the age/mbsf of both holes into an Excel spreadsheet to assess the regularity of sampling over time and changes in sediment rate, and got very messy results. But in any case I should have done this manually - this is not the reader's job.

I am sorry to be so pessimistic, but I am also convinced that if the authors want to reach a wider audience, they should submit their manuscript to a journal where people involved in biostratigraphy/taxonomy etc. are published.

Authors’ Response: We thank the reviewer for taking the time to assess our manuscript. We carefully read and answered the main concerns as follows:

We firmly believe that our manuscript is appropriate for Climate of the Past as it fits with the aims and scopes of the journal, as well as with the covered subject areas. Specifically, this work aimed at reconstructing paleoceanographic conditions within the Benguela Upwelling System by analyzing and interpreting changes in the planktonic foraminiferal assemblages. Thus, we used marine proxies (planktonic foraminifera) to assess ocean dynamics. Both ocean dynamics and marine proxies are subject areas in Climate of the Past. Furthermore, variations in the assessed paleoceanographic conditions could be linked to the existence of the Benguela Niño/Niña states in the South Atlantic Ocean during the Early to Late Pleistocene. In this respect, we believe that the detection of ENSO-related climatic conditions in older geological intervals (e.g., the Pleistocene) can be considered an important step forward in the understanding of the (paleo-)climatology of the region.

In this respect, several climate models suggested that ENSO conditions could occur since the Pliocene-Pleistocene and are strictly accompanied by strong zonal SST gradients (e.g., Rosell-Melé et al., 2014). Thus, it is scientifically valuable to understand if those climatic phenomena were present in the past geological intervals and if/why their duration can vary compared to the present day. We strongly emphasize that our manuscript represents a clear contribution to the knowledge on this topic of the paleoceanographic research community at large by providing evidence of Benguela Niño/Niña conditions based on the use of an important marine proxy, widely used for this scope.

Regarding the paleoceanographic aspect of the manuscript, we again believe that our work deeply contributed to obtaining valuable results in the field. All the main goals of the manuscript (see introduction, lines 54-59) were properly addressed as we were able to reconstruct the internal structure of the water column based on planktonic foraminifera. To the best of our knowledge, this approach provides data that cannot be gathered in any other way with the same level of detail.

Through the use of such paleobiological proxy data, we managed to detect important variations in the paleoceanographic conditions (thermocline fluctuations) in a very complex oceanic region, as well as to characterize the main water masses (Benguela Oceanic Current, South Atlantic Central Waters, and Indian Ocean water influxes), which exert an influence in the Benguela Upwelling System. Moreover, we statistically demonstrated that the increase in G. menardii within the BUS is related to the Agulhas Leakage and not to the variations in the depth of the thermocline. Thus, the analyses of the planktonic assemblages should be considered a valuable and highly reliable proxy to detect the intrusions of Agulhas waters in the SE Atlantic Ocean. Again, we emphasize that planktonic foraminifer assemblages are, to the best of our knowledge, the ONLY proxy data that can do so.

The time frame was clearly indicated throughout the manuscript, where we mentioned that the analyses of the planktonic assemblages cover the last 1.40 Ma. The chronostratigraphic framework was carefully evaluated using an integrated biostratigraphic approach and represented in figures 3-5 using the biostratigraphic zonations. We agree that age can also be inserted in the same figures to provide a better understanding to readers less familiar with geological data presented in the depth/thickness domain. For this reason, we will provide age intervals in Figure 3 and 5 as also recommended by reviewer 1. Figure 6 represents a conceptual model describing paleoceanographic conditions which are recurrent in our stratigraphic intervals. Thus, providing multiple age intervals in this figure can make it difficult to read. All the age information were provided in the two previous figures (3 and 5) exactly for this purpose.

We provided an extensive comparison of the detected paleoceanographic conditions with data presented in the literature (see lines 631-634; 672-676). Regarding the comparison of our data with previous SST reconstructions in a figure, we do not see merit in showing these datasets. The cited datasets are of high resolution and, therefore, are dominated by higher amplitude (Milankovitch) variability, often overshadowing potential long-term trends. Plotting them in their original resolution would, therefore, detract from the large-scale changes in mean state conditions that we try to define within this manuscript.

Furthermore, we also disagree that plotting already published SST data is in any way necessary or useful for our data interpretation and presentation – in fact, we feel it may even detract from the core messages we try to convey. After all, we are not only concerned with surface temperature changes but also evaluate the whole water column changes – something that, again, is impossible without the use of foraminiferal assemblage data-based proxy studies.

That being said, we still want to emphasize that high-resolution SST reconstructions were cited quite intentionally, as they were extremely useful in validating our results.

And yes, these data were, to a large extent, found through Google Scholar-based online research – a fact which we are quite honestly surprised that the reviewer seems to consider to be a bad thing.

Regarding the comment on sampling spacing: as the reviewer is obviously not familiar with any application, interpretation or methodology underlying proxy data that does not fall within their narrow scope of chosen research vocation, we will not further engage in such inflammatory and destructive comments. Suffice it to say that the chosen sample spacing of 1 sample per section (and not a regular time spacing!) was sufficient to provide a long-term record of the paleoceanographic changes in the region and, therefore, is very much appropriate for the proxy-based paleoclimate analysis we set out to present with this manuscript.

We also want to highlight that our manuscripts include biostratigraphic and quantitative data, statistical evaluations and paleoceanographic interpretations, which are unique and have a strong and well-founded historical basis for their interpretation and application in paleoclimate studies. These data cannot simply be gathered during a cruise, as the nature of data acquisition is uniquely time-consuming and labour-intensive – a fact that can only be justified by their undeniable value and unique usefulness as proxy data.

Thus, we find the (hopefully unintentional) insinuation that our manuscript is purely descriptive or similar to an IODP report downright destructive and of very poor scientific conduct. We consequently hope that such dismissive comments only stem from the reviewer's somewhat limited understanding of the wider field of proxy data methodology, application and interpretation previously noted by themselves. We consequently hope that our work and this scientific exchange may also serve as a possibility for the reviewer to broaden their scientific horizon beyond their own field.