Reply on RC1

As a general response to the comments regarding unclear statements, we agree that clear communication of the science is important. We will make a thorough effort in the revised manuscript to use accessible language and appreciate RC1 for highlighting sections that could be clarified. However, in some instances we have made the choice to reject comments calling for clarity as we feel the processes described are sufficiently well explained, albeit with discipline-specific language.

isotopic signal that is linked to the SASM. Released latent heat of condensation during the strong convection that characterizes the monsoon rainfall generates a Rossby wave response (Lenters and Cook, 1997). This establishes a strong correspondence between activity of the BH-NL system and the isotopic fraction that results from monsoon rainfall across the continent.
Unfortunately, the lack of low-frequency variability in the isotope-enabled climate models precludes a close examination of the causes of variability responsible for the observed variability of the first mode over time. However, our interpretation is consistent with the well-documented relationship between interhemispheric temperature gradients, latitudinal ITCZ displacements and SASM intensity changes on decadal to centennial timescales (Bird et

There is a lot more discussion and explanation needed on what exactly the authors did in the methods section. I found myself unable to follow the development of the statistical tests and the development of the pseudoproxy network. I read the papers the authors refer to here and am still confused as to what the authors in this paper did. I think the reader would be greatly aided if a clear summary can be presented in the methods section.
We will clarify the summary of the MCEOF analysis in the opening paragraph of section 2 (Data and Methods) and clarify the development of the pseudoproxy network. However, we are hesitant to add excessive detail to the methods given the length of the paper as currently written and due to the general method having been well described in a number of publications, all of which have been cited and discussed in the manuscript ( 3. The results and the discussion should be more clearly separated. In the results section, there is a lot of qualifying. Clear separation would make it easier to read.
Thank you for this suggestion. We will revise the results section to more clearly describing only the results of the analysis, while the interpretation of these results will be reserved for the discussion section. Please see more detailed discussion of this aspect in the line item comments below.

The rest of my concerns are laid out in line item format:
Line 25: The authors state: 'Model analyses suggests that the local isotopic composition is primarily a reflection of an upstream rainout processes.' I missed the full explanation of how the authors came to this conclusion based on the analysis. (It's possible I just misinterpreted something). However, a clearer explanation of this would be interesting.
We will clarify this conclusion to more clearly state that the results highlighting the role of upstream processes in the depletion of the isotopic records are derived from analysis of the Little Ice Age (LIA) and Medieval Climate Anomaly (MCA) period. We do not elaborate on the importance of the upstream rainout process as a mechanism driving the fractionation of oxygen isotope proxies as recorded in archives from within the SASM domain; rather we rely on the existing published literature that highlights this mechanism across the monsoon domain in both observations and isotope-enabled climate models (Grootes et al., 1989;Vimeux et   We wish to emphasize that the SASM is a highly seasonal phenomenon that is non-existent during the boreal fall and winter. A seasonal analysis therefore makes little sense given that the proxy records used in this study are driven to a first order by the climate climate signal within the monsoon wet season. This is precisely why this analysis, as is common practice, focuses exclusively on the mature monsoon season (DJF) as shown in Figure 1. Furthermore, both the original publications of the proxy records used in the network analysis (cited in Table 1)

Line 49: This first sentence is confusing.
We believe this sentence is fairly straightforward and maintains fidelity to discipline-specific language. Here we describe how the latent heat released during the phase transition from water vapor to cloud droplet leads to an anomalous heating of the mid/upper troposphere. This thermodynamic change leads to a dynamic response via the establishment of the Bolivian High -Nordeste Low system. We choose not to elaborate in too much detail on the fundamental atmospheric dynamics, but instead favor a focus on the isotopic variability and related processes in South America. The reader is also referred to key literature on this topic, such as Lenters and Cook (1997) and Chen et al. (1999). We therefore do not plan to change this sentence in the revised manuscript. Wong et al., 2021. This work argues for another driver on the SACZ and may be a good reason to do the kind of work that the authors outline here, good motivation for this study.

Introduction: I'm surprised that the authors didn't cite the work of
We appreciate being pointed to this paper, but after reviewing it, we find its focus is on the variability of the SACZ during the mid-late Holocene and thus concerns a different timescale than the decadal-to-centennial variability that is the focus of our manuscript. We also wish to reiterate that our analysis is entirely consistent with Novello et al. (2018), who explore a transect of δ¹⁸O records across the core of the SACZ and unequivocally show past variations in the location of the SACZ axis during the last two millennia.
Line 74: "The current interpretation of isotopic signatures in South American paleorecords..." The 'therefore' is used incorrectly. More importantly, I think it is delicate, but the current interpretations of the records from this region don't necessarily allow for the Rayleigh model, the records are interpreted assuming that model is true.
We will remove the word "therefore" from the sentence.
Our interpretation of Rayleigh distillation as the primary process driving isotopic fractionation of paleorecords from across the SASM domain draws on a wide range of literature as we have discussed in our previous responses. We nevertheless do appreciate that, in addition to the influence of upstream rainout and Rayleigh distillation, records are influenced by other processes, such as the amount effect in southeastern Brazil ( Line 123: The phrase "rule-of-thumb" has a difficult and potentially offensive history. I would encourage the authors to consider a different phrase. We understand the reviewers' concern about this phrase but fear it may be misplaced. The history of this expression is often misinterpreted as having an offensive history, while in reality no such connection exists (see: https://en.wikipedia.org/wiki/Rule_of_thumb). We nevertheless acknowledge that the misunderstanding around the etymology of this phrase seems to be common and therefore wish to avoid any unintentional association with the history some believe it invokes. We thus will modify the language to use the phrase "North's rule." Lines 130-135: It is difficult to understand what the authors did with the proxy records in order to complete the statistical assessments. What exactly does it mean for records to be "spliced together" What does it mean for the samples to be resampled to annual resolution (using what? how? which samples?). What does it mean for samples to be truncated (again, which ones?). If the results from this work are interpreted, than readers need to be able to follow how these proxy records were reconfigured, as it could impact the results.
We have made the code for this procedure publicly available for any reader wishing to understand the granularity of the process and/or replicate the process. Please note that section 2.1 is intended only to provide a brief overview of the statistical procedure, which is then described in much more detail in the later sections (2.2-2.4) for those readers interested in the specific details of the method.
We have revised the methods section to be clearer in response to the reviewer's comment. The relevant section of the manuscript is elaborated upon below for the specific questions that are raised by the reviewer.
We select fourteen oxygen isotope records, which all capture variability of the SASM. Using the age models and age model uncertainty of each of the fourteen records, we calculate ensembles of age models that contain 1,000 members using a Monte Carlo resampling. A unique oxygen isotope record is calculated for each age model ensemble member, generating a subsequent 1,000-member ensemble of oxygen isotope networks. To generate the modes of variability, an Empirical Orthogonal Function decomposition is calculated for each ensemble member of the oxygen isotope network and the median of the spatial and temporal modes are presented for analysis.
Due to the fact that not all records cover the entire period of study, we follow the methodology discussed in the published literature when merging discrete samples into composite records, referred to in our manuscript as "spliced together" records. We will clarify this language and more specifically refer to procedures from the published literature ( 'Resampling' and 'truncation,' refer to small processing steps we conducted for the oxygen isotope time series. The 'resampling' of samples to annual resolution is a statistical interpolation of the oxygen isotope data to a common resolution of one data point per year. This was done for the approximately annual sections of the Pumacocha record and the Pau D'Alho (ALHO6) sample that contain irregularly dated samples not corresponding to rounded calendar years. This procedure was also applied to those records which had a sampling resolution between two and seven years (PAR, PAL, P00-H1, MV, PIM4, DV2, SBE3+SMT5, TM0, Boto, JAR, and CR1). The expression 'truncation' indicates that we are only using data during the common period 850 -1850 CE. We will clarify these points in the revised manuscript. This truncation was necessary, as the MCEOF analysis requires that all records share a common time period.
Line 190: "Isotopic measurements excluded from previously published analyses were likewise excluded here (SBE3, PIM4), as were individual data points exceeding 3 standard deviations (MV30). Multiple isotopic measurements from a single depth were averaged to establish one value per depth (PAL3)." Do the parentheses in this sentence mean that these things were only done to the records in that parentheses? This is very difficult to follow. Furthermore, I'm concerned that these modifications weren't in the original publications of these records. If they weren't, then the authors should be very explicit about what and why they did these things.
We will modify this section to clarify that the analysis of raw samples is consistent with the analysis in the original publication of the data and instead refer the reader to the published literature. While we use the same procedure as in the original publications, we nonetheless do not use the published isotopic time series. We instead use the raw dates and raw isotopic data in order to establish new age models and corresponding ensembles of isotopic time series based on the Monte Carlo resampling techniques.
Our use of parenthesized sample names serves to specify which data sets have been modified by the stated procedure as was noted by R1. We find this to be consistent with currently published literature and to be grammatically correct. For MV, the gap is very small, approximately 4 years, or 0.4% of data, averaging across all the ensemble members for this record. Upon further review of the biascorrection, we find this step does not impact our result. It has been removed from the methods and both the code and manuscript will be updated to reflect this change.
Line 205: Do the authors mean that each of these speleothem records listed are composites? Or are all of these listed speleothem records combined for this analysis?
The records listed are composites. They are composited in the original publications and the raw samples are also composited for the purposes of this study. We name the records used in this analysis MV, JAR, PAR, SBE3+SMT5, PAL, and Boto, and the samples used in each of these composites is named in lines 124-129.
Section 2.5: I'm assuming (possibly naively) that the authors are using outputs from previous modeling efforts. If that's true, clarity on the point that these are previously published outputs would be useful. If I'm wrong (apologies) then greater detail on how these models were spun up and what biases may be included in them would be useful.
Thank you for this suggestion. The model output used in this research is from previously published research. We will make this more explicitly clear at the beginning of this section.
Line 224: "they rely on somewhat different forcing reconstructions." Please explain.
These forcing reconstructions are part of a suite of climate forcings developed for the Last Millennium simulations used in the Paleoclimate Modeling Intercomparison Project -Phase 3 (PMIP3) (Schmidt et al., 2011). The PMIP3 experimental setup includes a number of alternative climate forcing histories from which modeling groups could select particular forcings to be used in their own simulations. Removing the constraint that all modeling groups use the same climate forcings but imposing strict constraints on model configuration allowed PMIP3 simulations to test reconstruction uncertainty in a coherent modeling environment.
Because the discussion and comparison of different reconstructions in a coherent model environment is not the focus of this study, we will remove this phrase from the manuscript. We will retain the reference to the list of forcing datasets, which includes a justification for how the climate forcing database was constructed.

Line 243: "sub-grid scale processes" Like what? And how much does it impact the results? If it doesn't, why not?
In the model experiments used in this analysis, the horizontal resolution of the grid cells is ~ 2°, or ~ 222 km. Sub-grid scale processes refer to those processes operating on the spatial scale of less than one model grid cell. Because only a single averaged value is provided in the model output for a single grid cell, this averaged value obscures local conditions that function on smaller spatial scales than ~222 km. Some of these features are discussed in the introduction and include thermodynamic and microphysical processes such as condensation or dry air entrainment into clouds.
There are certainly sub-grid scale processes that are not well resolved in climate models. Indeed, Rojas et al. (2016) have documented that the PMIP ensemble for the LM simulates a circulation response consistent with the imposed LIA forcing over tropical South America, but that these circulation changes do not translate into precipitation changes, suggesting problems with implementation of feedbacks in the models or that the models may be too dependent on microphysics and convective parameterization schemes. We will provide more specific detail about these areas for model improvement in our revised manuscript.
However, it is worth noting that these sub-grid scale processes generally contribute to the noise that is unique to each paleorecord. Because the MCEOF extracts the fraction of the variance that is common to all records, synthesis of the coherent variability by the EOF analysis isolates the signal from the noise and we do not expect the key results to be impacted.

We will introduce the 'PPE' abbreviation when the concept is first mentioned in section 2.5.
Line 259: "a commonly employed proxy..." needs citation.
We disagree that this statement on the relationship between outgoing longwave radiation and tropical convection requires an additional citation. The citation provided for the dataset (Liebmann and Smith, 1996)  The magnitude of the dots for a given mode in Figure 2 represent the correlations between local proxy records and the Principle Component time series of that mode. This will be clarified when the figure is introduced in the text and in the caption of Figure 2. After some consideration, however, we prefer to retain the "same as in" construction for the sake of brevity. This is standard practice, significantly reduces the amount of text in the caption and the other text modifications will help clarify the figure content to the reader.

We will streamline the Figure 3 caption.
Line 280-282: An example of the needed separation between results and discussion: "This mode is interpreted as representative of the isotopic variability in the core monsoon region and is shown to vary on centennial timescales." Additionally, I would like to see more explanation on why the result is interpreted that way. Just not in the results.
We will rephrase this dynamic interpretation as a description of the results and expand further on the interpretation of the first mode in the discussion in section 4.1.1. Our interpretation of the first mode as representing the isotopic variability in the core monsoon region is because the highest loadings in EOF1 are associated with proxy records located in this core region. The centennial variability of the mode is apparent from PC1.

Line 296: There are two thoughts in this first sentence. Separate into what the results show and then the caveats.
The sentence will be modified to move the caveats to the Discussion, section 4.2.1.

Fig. 3: so the graphs on the far left, (f), (g) those are the pseudoproxy values?
Yes, Figure 3f,g are the pseudoproxy values. We will articulate this more clearly in the figure caption.
Section 3.2.2 I don't understand why these are being compared to each other. In my reading of the methods -the pseudoproxies are just the climate models with some white noise...so of course they agree well?
The pseudoproxy modes are derived from a discretely sampled number of locations corresponding to paleorecord sites that are also perturbed with noise. The agreement between the pseudoproxy modes and the spatially continuous climate model modes is not obvious a priori. We do not presume any coherency of the time series sampled from discrete points in the climate model space during the construction of the pseudoproxy time series. It is only after the EOF analysis of the network of these discrete points is performed that the modes emerge and are found to share similarities with both the proxy network and the spatially continuous EOF approach from the model data. The shared similarity of the modes of these three data sets confirms that the heterogeneous spatial distribution of the proxy network is able to capture a coherent signal across the domain that is shared by a heterogeneous (and continuous) dataset.
Line 384: I think it would help the reader if the authors explained the connection between OLR and rainfall. Then rainfall and depleted d18O, then depleted d18O and the stalagmite records.
We will provide a more detailed interpretation of the dynamic relationships in this analysis.

The anti-correlation between OLR and rainfall in the tropics is well established: as noted in the earlier comments, OLR is a commonly employed metric for deep tropical convection -more negative values of OLR indicate colder cloud-top
temperatures resulting from stronger convection. Stronger convection derives from enhanced vertical motion, which forces condensation of water vapor and thus, enhances rainfall.
The rainfall is related to changes in δ¹⁸O in the ways that have been previously described, primarily depletion via Rayleigh distillation along the moisture transport pathway in the upstream region of the monsoon (in our analysis, particularly upstream of the tropical Andes) and a mixture of path-dependent distillation and amount-effect driven distillation within the SACZ region. A variety of microphysical processes linked to tropical convective activity influence the downstream delta values of precipitation, however, these processes modify the fractionation along the moisture trajectory and rainout which primarily corresponds to the Rayleigh distillation model. Line 394: I don't quite see how the authors got to this conclusion: "SACZ activity within the dipole structure suggested by MCEOF2, and underscored by the precipitation dependence on OLR, is a function of the SAMS strength" based on the authors discussion of the data here. I see it in the results, but it's hard to follow how that connection is made in the text. This sentence will be clarified in the revised text.

The isotopic composition of rainfall is recorded in stalagmite records as it infiltrates karstic caves, dripwater degasses and calcite precipitates in
We also agree that the link between SACZ activity and monsoon strength is strongly supported by the results. We interpret the proxy-derived MCEOF2 as a dipole structure based on the anti-correlation between records to the northeast of the SACZ core (TM0, SBE3+SMT5, DV2, MV) and records to the southwest of the SACZ core (ALHO6 and CR1). We see this as clear evidence of a dipole within the decomposition of the isotopic records in the monsoon domain on the timescales of this study. In the below comment we describe in greater detail the links between the monsoon strength and dipole activity.
Line 410: "This was therefore a period when the SAMS was enhanced overall and both the ITCZ and the SACZ were displaced to the south of their mean locations" Again, I don't quite understand how this conclusion was derived following the in-text discussion.
In this sentence we have combined our results with the discussion of current literature, which documents a southerly displacement of the ITCZ during the LIA (Lechleitner et al., 2017;Steinmann et al., 2022). We will separate our results from this discussion of published literature in the revised manuscript.
Our results show a clear departure of the proxy-derived modes from their mean state. Both the departure seen in PC1 during the LIA and the dominant loadings over the core monsoon region are a clear indication of the enhancement of the SASM. The enhancement of the SACZ mode is similarly documented by the negative excursion of PC2 during the LIA together with the corresponding dipole of the proxy loadings in the SACZ region.
The paragraph starting on Line 414: OK… does this play a role? does it add support to your hypothesis? refute it? why is this paragraph summarzing these things here?
We highlighted these other factors influencing SACZ variability to provide a more complete perspective, but we agree that much of this discussion is beyond the scope of this study. We will remove this paragraph in the revised manuscript.
Line 430: "Those records are more sensitive to large-scale circulation changes and related non-monsoonal influences outside the mature monsoon season (DJF)." What is this based on?
Records located at sites distal from the core of the monsoon domain are more sensitive to influences from other systems that border the monsoon domain, thereby providing potential pathways for moisture from alternative sources other than the SASM to reach these proxy sites.  Figure 5 is primarily focused on the significance of upstream convection relative to the changes of the δ¹⁸O values measured in the isotopic proxy records during the MCA and LIA. The large-scale changes in upstream convective activity seen during the MCA and LIA in the models is consistent with our MCEOF1/EOF1 pattern representing the core monsoon circulation and the corresponding changes in strength present in PC1.

Our interpretation of
The lack of broad statistical significance in the modeled isotopic and precipitation changes is consistent with our discussion of the challenges models face in responding to external forcings. We also clearly state in our interpretation that isotopic variability is not primarily a response to an amount effect but rather a response to changes in upstream 500 hPa vertical motion. Thus we do not expect to see co-location of significance between δ¹⁸O and precipitation anomalies.

Done.
Line 481: I agree that topography presents a challenge to models, but I don't understand why this site (Boto) has trouble and the other sites along the Andes do not. Or should we question the correlations at any of the sites along the Andes?
We may not have correctly phrased this, but we wish to point out that the challenge in correctly simulating the climatic controls over the Boto site is not primarily a topographical argument. Modern climatological studies show that different modes of SASM precipitation can produce significant spatial variability between the northern and southern part of the central Andes at interannual timescales (Vuille and Keimig, 2004  Also, please note that this regional anomaly is limited to the MCA period only (Fig. 5), as proxy data and model simulations are in agreement during the LIA. Hence, we believe that the change in moisture transport to this distal site during the MCA constitutes the main challenge for the climate models. We do not call into question the model representation at the other Andean sites, given that the challenge is not primarily due to topography, but rather related to the interactions between tropical and extratropical climate that influence this transition region of the SASM. Indeed, during both MCA and LIA the other Andean records show good agreement with the model data. This will be clarified in the revised manuscript.