Thanks to the authors and editor for the opportunity to review this paper. I see many potential policy outcomes from this work, and look forward to seeing it published. As it stands, I think there are some outstanding issues which need further clarification. Please see the attached for my detailed review. 

Review „Spatiotemporal ITCZ dynamics during the last three millennia in Northeastern Brazil and related impacts in modern human history“ by Utida et al.

I have read with great interest the discussion paper by Utida et al. The authors analyze spatiotemporal ITCZ dynamics during the last three millennia in NE Brazil (NEB), and claim to relate their inferences to modern human history. The study presents partially replicated speleothem proxy records from two caves in NE Brazil, and provide an overview of past (hydro)climate trends and variability in the greater region of NEB of the southern margin of the ITCZ.

This new data set is sound and I have no doubt about the quality of the applied methods and presented data. In principle, the scientific significance is valid, since this dataset complements the northern South American speleothem record in high resolution. However, I have some concerns about the structure and clarity of the manuscript, which I feel needs some improvements before final publication.

Main comments:

• The structure:

I find most of the conclusions concerning ITCZ dynamics intriguing and interesting. However, I found the manuscript sometimes hard to read and some parts of the discussion are not easy to follow. For example, the section in L388ff has a rather unclear structure. The first half seems to be organized chronologically along the time of record, and describes the observed trends. But before this discussion is finished, the discussion jumps to comparing relationships between proxies, and refers to sections of the record which have not been described yet. Later on, the discussion also jumps from describing potential processes back to certain events and forth to other aspects again. I feel like the whole discussion should be carefully restructured and streamlined to build the arguments better on each other, and to provide the reader a common thread throughout the manuscript to prepare and justify your conclusions properly. I suggest to choose a consistent, logic structure, such as building up the discussion more strictly chronologically along your record, and also discuss trends first and events later separately?! Another possibility would be to bring the proxy interpretation first, and then compare to other records and discuss the forcings and consequences… There are several possibilities, but please do not mix it all up…

I also strongly suggest to put special effort in elaborating how the two parts of the discussion (i.e., the paleo-record description, and the discussion of historical droughts) actually build on each other, and better justify why both aspects need to be discussed in one paper. In the current version of the manuscript these appear more as two separated stories.

• U/Th Results description

I miss a proper description of the U/Th results in the main text. This should e.g., comprise U and Th concentrations, uncertainties, Th contamination, description of inversions, etc, … (check Dutton et al. 2017 as a guideline to report U-series data). This is also important due to the presence of both calcite and aragonite, where we would expect an influence on the ages if recrystallization occurred! In addition, a statement concerning the final uncertainties of the age-depth model is essential, also regarding the several outliers. This is particularly relevant when reporting absolute ages for extreme events! From the so presented age models, it is not at all clear if the dating supports an annual precision of a single drought event, or the unequivocal allocation to an event reported in the historic record.

I have some more general comments to the style of the writing and presentation, which I summarize here. Please find specific locations related to the following points in my minor comments along the text:

• Across the manuscript I found repetitive statements, but also rather irrelevant information. This makes the reader lose focus, so I suggest to try to shorten/streamline the text in general.
• In many figures, some aspects are hardly visible. Please improve accessibility, e.g., text sizes, increase size of markers of locations, use colors that are better visible.
• Sometimes past and present tense is mixed, please check language style.

Minor comments:

L49 weakening

L62-63: Is there a reference for this statement?

L91: I think the Lechleitner Paper is from 2017.

L129-131 is this relevant?

L138: Any idea why the cave temperature is considerably lower than the annual mean temperature? Is this relevant for your data?

L148: I feel like most of this section is rather results than material/methods description?

L149ff: There is a lot of discussion of the different sectors within NEB, it may be helpful for the discussion and the readers to indicate those in a figure?

L164: How is “most significant” defined?

L174: “is primarily the result of a shorter rainy season”. This is not quite what is described above. There you write that the rainy season has the same length but is weaker?

L175: “The anomalous length…” See previous comment, according to your own results, this is only for the wetter years.

L189ff: If this is relevant for the discussion later, I feel like the authors should clearly define the difference between ITCZ related rainfall in NEB, and SASM related rainfall in S-NEB. Some reader may not be able to recall the exact difference at once…

L229-230: This information is not relevant for this study.

L282 to significantly reduce

L327: Avoid repetition of the method description.

L343: Why is a composite record only constructed for δ¹⁸O? what about the early phase 2-3k BCE? It is shown in the figure, but I didn’t find a statement why it is shown but not discussed

L389: if the age axis is correct, the oldest period of the RN composite is around 1.5 to 1k BCE?

L390: I see rather positive anomalies between 1 and 0.5 BCE…

L391ff: Confusing section, please clarify. In the previous sentence you state, there is soil erosion, here you state that did not contribute much… Now what?

L397: more negative as compared to what? to me, the δ¹⁸O values are rather higher than in the earlier part of the record… please clarify.

L399ff: This paragraph is hard to follow. Please don’t start compare/discussing sections of the record that haven’t been mentioned before… (here the LIA suddenly pops up)

L408: On the millennial scale, yes… since you also mention shorter timescales earlier, I would clarify this here…

L413: Unclear, why is this?

L414 do you mean “latter”?

L415: Very vague statement, please specify. Also, how well are the lake sediments dated and is that comparable to your chronology?

L420 Maybe indicate the insolation curve in the Figure?

L421 persistently (?)

L426ff: This is an interesting conclusion which is however barely discussed beforehand. The discussion here rather ends quite abruptly. I feel this could be more elaborated, because it seems to relate to the statement in the abstract, that you can make inferences on spatio-temporal ITCZ variability?

L432: very vague and unclear which characteristics are meant

L429ff (the whole section) difficult to follow here, you jump from describing a trend to single events, and then to processes again - not clear where this leads to? please provide the reader with some kind of guidance in between, maybe in form of a summary and/or statement which observation will be tested/explained now…

L437: unclear what information your record adds to this aspect, and how this relates to the discussion?

L441: now the discussion jumps again back in time to another event… I would bring this example later to showcase a potential relationship to Atlantic temperatures…?

L447: I suggest to turn the argument around - the idea is that ITCZ displacements are forced by temperatures, so we check if there is a relationship of our record to AMV?

L448ff This sentence seems incomplete

L452: Have you checked other records of AMV / Atlantic SSTs that allow to check if the Lapointe record is representative for the entire basin during these times or not?

L462ff As far as I understand the plot, there is no PDV record in the plot, so how do you infer a cold phase of PDV during that time? I guess you refer rater to Fig 5, but still I suggest to explain which record / curve you are referring to here exactly and what they are showing? Is the Pacific SST gradient a measure of PDV? this curve shows centennial scale variability, but not at all decadal?

L494ff: what does the δ¹³C record tell in this time? extreme events could be also visible there, the record looks quite “spikey”

L495: How do the other speleothem records compare during this time? I understand that they have lower resolutions, but to support your point and strengthen your arguments (e.g. concerning age model uncertainties, etc) a zoom into the comparison of the different proxy records might be helpful? Also, how would ISCAM move the TRA7 record with respect to the original agemodel? This gives also a hint regarding dating uncertainty…

L497ff: statements like this require a proper report of dating and age model uncertainties. From visual inspection there are some ages which have quite high uncertainties, which could limit the fidelity of such a record to absolutely date extreme events with annual precision! It could be also short-term hiatuses, that last longer than a single year…? I understand that the TRA7 age model is part of another paper, but then please still give a statement here, because this is relevant for your conclusions. It is also not clearly visible from the plot of the age model in the supplement.

L538ff: how many droughts are not recorded in your stalagmite record? the reference is not accessible, so please provide a clear statement, or, better, a plot/histogram of all droughts reported by the other study in Fig 6

L551: Discussion ends quite abruptly, following from your section 5.1 one would at least expect a hypothesis of a forcing mechanism of the drought occurrence? How is the drought frequency related to what you found out from your record of the past 2.5ka? I suggest to elaborate this a little bit more…

Figures

Figure 1: Locations hardly visible, please increase the size of the text and the stars. Also No. 5 is barely visible, please choose other colors.

Figure 2: Increase symbols for locations. Please improve visibility in general. Caption should be streamlined, “precipitation amount” is mentioned twice in the first sentence (L199-201). Correlation maps is repeated in L199 and L204. GNIP is repeated in L200 and L207. No need to repeat all information to all caves again, it is also ok to refer to the previous figure…

Figure 3: Please check if colors are color-blind friendly (red and green mixed…?) Also, why is the early phase of TRA7 between 3 and 2k not included in the composite?

Supplementary material

Tables S1, S2, S3: Please check decimal and 1000s delimiter, there are different styles used (comma and points mixed, sometimes comma as 1000s delimiter, sometimes not). Also “delta”234U instead of d234U.

Figure S4: Any ideas for the outliers, e.g., in TRA7 or FN1? Also, why is the age model of FN1 systematically older than the stalagmite ages? Also, why  do you show ISCAM uncertainties, but COPRA average age model? Why not show ISCAM and COPRA in comparison?

References

Dutton, A., Rubin, K., McLean, N., Bowring, J., Bard, E., Edwards, R. L., ... & Yokoyama, Y. (2017). Data reporting standards for publication of U-series data for geochronology and timescale assessment in the earth sciences.Quaternary Geochronology, 39, 142-149.