Enhanced subglacial discharge from Antarctica during meltwater pulse 1A

Subglacial discharge from the Antarctic Ice Sheet (AIS) likely played a crucial role in the loss of the ice sheet and the subsequent rise in sea level during the last deglaciation. However, no direct proxy is currently available to document subglacial discharge from the AIS, which leaves significant gaps in our understanding of the complex interactions between subglacial discharge and ice-sheet stability. Here we present deep-sea coral 234U/238U records from the Drake Passage in the Southern Ocean to track subglacial discharge from the AIS. Our findings reveal distinctively higher seawater 234U/238U values from 15,400 to 14,000 years ago, corresponding to the period of the highest iceberg-rafted debris flux and the occurrence of the meltwater pulse 1A event. This correlation suggests a causal link between enhanced subglacial discharge, synchronous retreat of the AIS, and the rapid rise in sea levels. The enhanced subglacial discharge and subsequent AIS retreat appear to have been preconditioned by a stronger and warmer Circumpolar Deep Water, thus underscoring the critical role of oceanic heat in driving major ice-sheet retreat.


Overview
The manuscript by Li and others reports a study of coral U-series data from the Southern Ocean that identifies a rise in the δ 234 U of seawater during deglaciation.There is a distinct pulse overlapping in time with meltwater pulse 1A, that the authors convincingly relate to recent/nearby discharge from the Antarctic Ice Sheet.
My expertise lies in the realm of U-series geochronology and geochemistry, so much of my feedback focuses on the reporting, screening, and interpretation of this data.While the tracer particle simulations appear robust, I am not able to rigorously evaluate them, and I trust that another reviewer has been able to do so.
Overall, I find the study's results and conclusions compelling and the methods rigorous.However, I would like the authors to more explicitly and quantitatively describe the nuances and complexities of the coral U-series systematics and how they handle them.Clearer and more detailed descriptions of their approach will not only fortify their interpretations but also benefit a broad reading audience of both specialists and non-specialists, alike.To this end, I detail a few general comments, followed by specific line-by-line comments.
I recommend this article for publication in Nature Communications so long as the following comments are adequately addressed.

Respectfully, Graham Edwards
General Comments: • The authors repeatedly acknowledge that the U-series system is susceptible to minor perturbations (e.g.lines  and dismiss the minor variability in the context of the larger trends of the data.I agree with their conclusions, but I think the manuscript would benefit from a very explicit statement that the study benefits from its dataset's large population, which allows the authors to identify trends within the substantial and significant (∼3 ) heterogeneity and scatter of the data.The authors should compare the scale of perturbations with the precision of their time-δ 234 U i record and interpretations thereof.(see comment on line 105-114) • The manuscript uses abbreviations for sample sites extensively.The clarity of the paper would benefit from a tabulated summary of these.I suggest a legend for Fig. 1 would provide readers with an efficient reference for the abbreviations and corresponding site locations.
• There are a handful of typos and grammatical errors throughout the manuscript.I have made a few suggestions to clarify confusing phrasing, but I encourage the authors to do a close proofreading prior to publication.The referenced study (Blackburn+ 2020) reports data from subglacial chemical precipitates but not (necessarily) subglacial lake carbonates.While there are U-series data from Mc-Murdo Dry Valley lake carbonates presented in that study, these lake carbonates derive from proglacial, rather than subglacial, lake environments (e.g.Higgins+ 2000, https: //www.jstor.org/stable/521000).The subglacial precipitates may have formed in subglacial lake environments, but that is not conclusive.

73
"has a somewhat muted influence on the δ 234 U i values" -this assertion and Supplementary Figure 1 should be described in more detail.A more quantitative treatment of this is necessary to establish how insensitive δ 234 U i is to the 230 Th correction.Specifically, I think a more thorough discussion of Supp.Fig. 1 in the methods or in the figure caption (see comment on Supp.Fig. 1 below) would help to elucidate the authors' point.

83-90
This list is very important but the sentence is long and complex.I suggest that authors revise this sentence for clarity and, in particular, reference the specific groupings in Fig. 1 to help readers visualize sample sites.
105-14 This is a very important and well-articulated discussion on the inherent variability of coral U-series records.In lines 113-4, the authors state that the influence on large-scale trends is minor.I agree with this statement, but I would like the authors to more quantitatively discuss the relationship between this variability and their results/interpretations.In particular, how does this variability compare to the uncertainty about the trend (i.e.shaded regions in fig.2).This relates closely to the bulleted comment above: the large dataset allows the authors to resolve a higher signal to noise in the compilation compared to individual samples.Stating this explicitly will strengthen the implied utility of this dataset and the authors' approach.

123-132
This decay-only model overlooks that there is always an input of excess 234 U into the ocean (hence the δ 234 U i ∼144 at >20 ka and 146.8 in modern ocean water).The authors should numerically model the δ 234 U evolution with an input flux of 234 U that maintains LGM δ 234 U compositions, and examine whether the loss of 234 U is still faster than the decline in the δ 234 U i record.I did a cursory attempt at this calculation: if I impose a flux of 234 U sufficient to sustain the system at δ 234 U=144 indefinitely and start with a δ 234 U=148.9 , the δ 234 U of the system decreases by <1 over 12 ka.To me, this does not suggest that there were additional pulses of high-234 U waters.Rather, the gradual decline of δ 234 U over the Holocene reflects the gradual mixing of 234 U-enriched seawater with lower-δ 234 U water masses.I encourage the authors to perform these calculations with their own derivations (lest I made a mistake in mine) to validate these results and adjust their discussion accordingly.Th content on "error."Is there an unstated uncertainty associated with the value of 5000, does the correction end up incorporating uncertainty from an additional isotope measurement that scales with 232 Th, or are the authors reporting error as the offset between uncorrected and Th-corrected values?The authors should explicitly state how the described "error" arises.
373-8 While I concur that the internal consistency of the data are not affected by the use of outdated decay constants, the authors should also discuss the effect of these methods on the external reproducibility of ages, since they compare the timeseries with other paleoclimate records (e.g.Fig. 4).While I suspect the differences are minor, this systematic error should be quantified like it is for δ 234 U i .Additionally, per the recommendations of Dutton+ (2017, https://doi.org/10.1016/j.quageo.2017.03.001), the authors should state the decay constants used to calibrate standards and calculate ages in the caption or in a footnote to Supplementary Table 1.
393-4 I believe this 3 threshold is derived from the upper limits of variability in Supp.Figs 4 and 5.The authors should reference these to identify the origin of the threshold.

Supplementary Information
Supp.Fig. 1 What do the dashed red lines indicate?Please provide units on the y-axes (years and , presumably?).The authors also need to clarify how they calculated the reported and modeled errors.As in the comments for lines 73 and 361-4, it is not clear how the "error" values reported in this figure were calculated, obscuring confident interpretation of the figure.The authors use "error" and "uncertainty" seemingly interchangeably in the caption, which further challenges interpretation.Without more clarity, this figure fails to adequately substantiate the data screening criteria used in this study.
Supp.Fig. 3 Please describe the black curve and gray band.Presumably this is the filtered timeseries from Figs. 2 and 4, but this should be stated explicitly.
Supp.Fig. 4 Provide information for which samples were measured in duplicate so that readers may cross-reference with the table.
Supp Fig. 7 Provide legend labels for the green and yellow curves or describe in caption.
Reviewer #2 (Remarks to the Author): Review of the paper by Li et al. entitled as "Enhanced subglacial discharge from Antarctica during meltwater pulse 1A".
Li et al. presented the results of uranium series nuclides measurements on deep sea corals obtained from the Southern Ocean.They found distinctive changes in del234Ui in seawater during the last deglaciation.The changes are attributed as deglacial climate events sicne they are likely coincided with other paleoclimate data including sea level changes.In particular, sharp rise in del234Ui occurred at the time when global sea level rose rapidly known as Mwp1A (melt water pulse 1A).del234Ui is unique proxy of chemical weathering and authors' data are scientifically sounds.Hence, I would like to see the paper be published eventually though I think their discussions need to be modified before accepting it for publication.
The uniqueness of this dataset is that their age model is based on radiometric dates based on uranium series dating.This is very strong point of the data since it is often difficult for paleoceanographic observations obtained from deep sea sediments in the Southern Ocean due to large reservoir effects as well as dilution effects from dead carbon released from Antarctic continent to radiocarbon age models.Hence, I disagree with authors to compare Iceberg rafted debris (IBRD) flux to discuss paleoenvironmental conditions because of two reasons.The first is that IBRD has many peaks and their age model is based on correlations to the ice core data under several conditions.Thus, their robustness of the age models for millennial scale changes are very weak.The second, IBRDs are detrital materials remained in deep sea sediments without knowing provenances.No isotope data on IBRD were conducted such Nd and Pb isotopes to identify origins of IBRDs.Therefore, using IBRDs data to discuss Antarctic ice sheet melting history is not convincing.Instead, they should use other studies reconstructing ice sheet history as well as paleoceanography of the Southern Ocean.
Previous Antarctic ice sheet melting history reconstructions include exposure dating using cosmogenic nuclides (eg., Yamane et al., 2010;Johnson et al., 2014;2020), marine sediment core data using compound specific radiocarbon dating (eg., Johnson et al., 2021;Yamane et al., 2014) and ice core data (eg., Cuffey et al., 2016;Das and Alley, 2008).Further, since del234Ui is a strong indicator of chemical weathering, authors should consider comparing their results with reactive phase 10Be/9Be in sediments on and offshore of Antarctica (eg., Sproson et al., 2021Sproson et al., , 2022;;Behrens et al., 2022;White et al., 2019).Changes in basal condition provide large consequences of ice sheets behavior and are important constraints on ice sheet models.Studies employing 10Be/9Be in sediments also pointing towards Atmosphere and Ocean interaction related Antarctic ice melting in the past that is in line with currently employed by authors (eg., Yokoyama et al., 2016 SOM;Sproson et al., 2022).Hence, they should discuss their data with above mentioned and beyond data in comprehensive manner to make their data very important.
Because global sea level data from low latitude are reconstructed using U-series dates (Deschamps et al., 2012;Yokoyama et al., 2018), it is useful to expand the section more on comparison between global mean sea level studies.Authors did it for some degrees on Mwp1a, but it would be good to see more focusing on the topic of source(s) of meltwater if it is coming from only Northern hemisphere ice sheets or not (Brendryen et al., 2020).Geophysical fingerprint techniques (eg., Yokoyama and Purcell, 2021) suggested Antarctic involvements (eg., Clark et al., 2002;Deschamps et al., 2012) but far-field fingerprint technique is indirect method so that the current data would provide more constrain on those scenarios combining on authors' previously published dataset (Chen et al., 2016).
Finally, just a minor point but I would not describe U-series dates as absolute dates.Any radiometric dates are based on several assumptions including closed system of nuclides, correctness of decay constant, pristine nature of samples and others.Hence, please change the term to other appropriate one.
Review text in black, authors' replies in blue.

Replies to Reviewer #1
The manuscript by Li and others reports a study of coral U-series data from the Southern Ocean that identifies a rise in the δ 234 U of seawater during deglaciation.There is a distinct pulse overlapping in time with meltwater pulse 1A, that the authors convincingly relate to recent/nearby discharge from the Antarctic Ice Sheet.
My expertise lies in the realm of U-series geochronology and geochemistry, so much of my feedback focuses on the reporting, screening, and interpretation of this data.While the tracer particle simulations appear robust, I am not able to rigorously evaluate them, and I trust that another reviewer has been able to do so.
Overall, I find the study's results and conclusions compelling and the methods rigorous.
However, I would like the authors to more explicitly and quantitatively describe the nuances and complexities of the coral U-series systematics and how they handle them.Clearer and more detailed descriptions of their approach will not only fortify their interpretations but also benefit a broad reading audience of both specialists and non-specialists, alike.To this end, I detail a few general comments, followed by specific line-by-line comments.
I recommend this article for publication in Nature Communications so long as the following comments are adequately addressed.
Reply: We appreciate the reviewer's positive assessment of our manuscript and concur with the suggestion regarding the need for a more comprehensive description of the coral U-series systematics in the revised manuscript.We have thoroughly reviewed all the comments provided below.

General Comments:
The authors repeatedly acknowledge that the U-series system is susceptible to minor perturbations (e.g.lines 57-77, 105-14) and dismiss the minor variability in the context of the larger trends of the data.I agree with their conclusions, but I think the manuscript would benefit from a very explicit statement that the study benefits from its dataset's large population, which allows the authors to identify trends within the substantial and significant (∼3‰) heterogeneity and scatter of the data.The authors should compare the scale of perturbations with the precision of their time-δ 234 Ui record and interpretations thereof.(see comment on line 105-114) Reply: We appreciate the reviewer's emphasis on the importance of explicitly conveying how our work benefits from the high sample, particularly concerning the interpretation of deep-sea coral δ 234 U data from the Southern Ocean.In response to this concern, we conducted Monte Carlo simulations and utilized the Wilcoxon rank sum test to examine whether the transient perturbations observed at 16 -14 ka can be attributed to the internal variability of coral δ 234 U (Supplementary Figure 6).
To perform this analysis, we generated 10,000 synthetic time series of δ 234 U by introducing errors to δ 234 U measurements for different groups of corals (Fig. 3b).These errors were represented as normally distributed random numbers with a mean of zero and specified standard deviations.Subsequently, we calculated p-values, reflecting two-sided Wilcoxon rank sum tests comparing the δ 234 U medians between Group Ⅰ and Group Ⅲ, as well as between Group Ⅰ and Group Ⅱ, for each time series.Our results unequivocally demonstrate that the perturbations observed in Group Ⅰ can be statistically distinguished from those in contemporaneous seawater (Group Ⅱ and Ⅲ) at 16 -14 ka, given a standard deviation of 2‰ for coral δ 234 U. Considering that the internal δ 234 U variability, as determined by the late Holocene (< 1 ka) samples, is ~1.3‰ (Supplementary Figure 5), it becomes evident that the extensive coral population within the targeted age interval of 16 -14 ka is more than adequate to differentiate the elevated δ 234 U spikes from those of contemporaneous seawater (Fig. 3).We have thoroughly revised the discussion section and the supplementary information accordingly.
The manuscript uses abbreviations for sample sites extensively.The clarity of the paper would benefit from a tabulated summary of these.I suggest a legend for Fig. 1 would provide readers with an efficient reference for the abbreviations and corresponding site locations.
Reply: Thanks for pointing this out.We have revised the manuscript to reduce the frequency of abbreviations used for the sample sites.
There are a handful of typos and grammatical errors throughout the manuscript.I have made a few suggestions to clarify confusing phrasing, but I encourage the authors to do a close proofreading prior to publication.
Reply: We have carefully checked the words and grammar thoroughly.

Comments by line #:
Line #39 The referenced study (Blackburn+ 2020) reports data from subglacial chemical precipitates but not (necessarily) subglacial lake carbonates.While there are U-series data from Mc-Murdo Dry Valley lake carbonates presented in that study, these lake carbonates derive from proglacial, rather than subglacial, lake environments (e.g.Higgins+ 2000, https: //www.jstor.org/stable/521000).The subglacial precipitates may have formed in sub-glacial lake environments, but that is not conclusive.
Reply: We agree with the reviewer that the usage of lake carbonates is ambiguous here since subglacial chemical precipitates are not necessarily formed in the lake environment and the form of the chemical precipitates includes both carbonate minerals and amorphous silica/opal precipitation.We have revised this sentence "High δ 234 U values of a similar magnitude have also been observed in chemical precipitates formed in subglacial aquatic environments in East Antarctica".
Line #73 "has a somewhat muted influence on the δ 234 Ui values" -this assertion and Supplementary Figure 1 should be described in more detail.A more quantitative treatment of this is necessary to establish how insensitive δ 234 Ui is to the 230 Th correction.Specifically, I think a more thorough discussion of Supp.The modeled uncertainties were calculated by considering a relatively large uncertainty in the initial 230 Th/ 232 Th atomic ratio (2 × 10 -4 , 2σ) (Bradtmiller et al., 2009) and were propagated with a Monte Carlo method when solving the age equation (Burke and Robinson, 2012;Chen et al., 2015;Li et al., 2020).We have revised the caption of Supplementary Figure 1 to clarify this.

Lines #83-90
This list is very important but the sentence is long and complex.I suggest that authors revise this sentence for clarity and, in particular, reference the specific groupings in Fig. 1 to help readers visualize sample sites.
Reply: Thanks for pointing out this.We have revised this paragraph as well as Figs. 1 and 2 to clarify this.
Lines #105-114 This is a very important and well-articulated discussion on the inherent variability of coral U-series records.In lines 113-4, the authors state that the influence on largescale trends is minor.I agree with this statement, but I would like the authors to more quantitatively discuss the relationship between this variability and their results/interpretations.
In particular, how does this variability compare to the uncertainty about the trend (i.e.shaded regions in fig.2).This relates closely to the bulleted comment above: the large dataset allows the authors to resolve a higher signal to noise in the compilation compared to individual samples.Stating this explicitly will strengthen the implied utility of this dataset and the authors' approach.
Reply: Please see our response to the first general comment.

Lines #123-132
This decay-only model overlooks that there is always an input of excess 234 U into the ocean (hence the δ 234 Ui∼144‰ at >20 ka and 146.8‰ in modern ocean water).The authors should numerically model the δ 234 U evolution with an input flux of 234 U that maintains LGM δ 234 U compositions, and examine whether the loss of 234 U is still faster than the decline in the δ 234 Ui record.I did a cursory attempt at this calculation: if I impose a flux of 234 U sufficient to sustain the system at δ 234 U=144‰ indefinitely and start with a δ 234 U=148.9‰, the δ 234 U of the system decreases by <1‰ over 12 ka.To me, this does not suggest that there were additional pulses of high-234 U waters.Rather, the gradual decline of δ 234 U over the Holocene reflects the gradual mixing of 234 U-enriched seawater with lower-δ 234 U water masses.I encourage the authors to perform these calculations with their own derivations (lest I made a mistake in mine) to validate these results and adjust their discussion accordingly.
Reply: We acknowledge the reviewer's valid point that the previous decay-only model employed in our study did not account for the constant riverine input of excess 234 U into the ocean.Furthermore, we agree that the gradual decline of δ 234 U observed over the Holocene likely reflects the gradual mixing of 234 U-enriched seawater with lower-δ 234 U water masses.
To address this, we have added this point into the discussion.As the reviewer suggested, we have checked the calculations with our own simple model, which we elaborate on here.This is a simple calculation that incorporates both 234 U decay and the riverine input of excess 234 U. We assessed the necessary changes in riverine δ 234 U and riverine U flux relative to the last glacial period to explain the observed 2‰ drop in seawater δ 234 U over the Holocene (Response figure).
In this model, we assumed that seawater δ 234 U (144‰) was in a steady state at the end of the last glacial period (30 -18 ka).This assumption requires a riverine input of 347‰, taking into account a U residence time of 400 kyr (Dunk et al., 2002;Henderson, 2002).steady state at the end of the last glacial period (30 -18 ka).This assumption requires a riverine input of 347‰, taking into account a U residence time of 400 kyr (Dunk et al., 2002;Henderson, 2002).For the last deglaciation (18-11.5 ka), we employed two modeling scenarios (red lines): one involving an increase in the δ 234 U of riverine input while keeping the U flux the same (a) and the other involving an increase in U flux while maintaining a constant δ 234 U of 347‰ (b).
These adjustments were made to align the model outcomes with the deep-sea coral data (grey lines enveloped by shading).In the context of the Holocene period, we considered two 16 -14 ka.This test yielded a p-value less than 10 -3 , even when considering the internal variability of coral δ 234 U (Fig. 3 and Supplementary Figure 6).This statistically significant result demonstrates that the δ 234 Ui excursion (Group Ⅰ) can be distinguished from the largescale trends (Group Ⅱ and Ⅲ).The main text has been revised to reflect these changes.
Lines #143-145 This sentence is not clear to me.As written, it seems to suggest that the absence of high δ 234 U in the Pacific and Indian oceans during HS1 accounts for the δ 234 U spike in the Southern Ocean.Please either elaborate or rephrase for clarity.
Reply: Thank you for bringing this to our attention.We have revised the sentence and relocated it to the section where we discuss the potential mixing of seawater from other ocean basins.
Lines #147-149 For clarity, reference Fig. 2C and state that the deep SS water does not show this.
Reply: Suggestion accepted.We have revised this sentence to clarify it.
Fig. 1 in the methods or in the figure caption (see comment on Supp.Fig. 1 below) would help to elucidate the authors' point.Reply: The influence of initial 230 Th correction on final age and δ 234 Ui uncertainties were evaluated by calculating the uncertainties of age and δ 234 Ui for a modern sample (age = 0 year, δ 234 U = 146.8‰)and a 20000-year-old sample (δ 234 Ui = 146.8‰)with varying 232 Th contents.
For the last deglaciation (18-11.5 ka), we employed two modeling scenarios: one involving an increase in the δ 234 U of riverine input while keeping the U flux the same and the other involving an increase in U flux while maintaining a constant δ 234 U of 347‰ (red lines).These adjustments were made to align the model outcomes with the deep-sea coral data (grey lines enveloped by shading).In the context of the Holocene period, we considered two scenarios.Scenario 1 (solid line) demonstrates how changes in seawater δ 234 U can be achieved by decreasing either δ 234 U of riverine input or U flux relative to the last glacial period to align with the deep-sea coral data over the Holocene while scenario 2 (dashed line) reflects the results when either δ 234 U of riverine input or U flux remains consistent with the values of the last glacial period.The model results indicate that either a ~80‰ drop in δ 234 U of riverine input or a reduction of more than half in riverine U flux compared to the last glacial period is necessary to account for the Holocene decrease in seawater δ 234 U of ~2‰.Of course a minor combination of both of these factors, as well as mixing with lower δ 234 U waters could also be consistent with the data.Given the under-constrained nature of the calculation and its multiple controls, including uranium mass vs isotopic flux and ocean mixing assumptions, we think a robust exploration of this is beyond the scope of the current study.We have thus removed the decay calculation from the previous figure and opt not to include the figure made for the review response below.We have amended the text as described above in line with the reviewer's important insight on the multiple controls on the Holocene decline in δ 234 U and we hope to explore this further in future work.Response Figure | The model results depicting relative changes in seawater δ 234 U during the last deglaciation and Holocene periods in comparison to the deep-sea coral δ 234 Ui record from the Southern Ocean.This model assumes that seawater δ 234 U (144‰) was in a scenarios.Scenario 1 (solid red line) demonstrates how changes in seawater δ 234 U can be achieved by decreasing either δ 234 U of riverine input (a) or U flux (b) relative to the last glacial period to align with the deep-sea coral data over the Holocene while scenario 2 (dashed red line) reflects the results when either δ 234 U of riverine input (a) or U flux (b) remains consistent with the values of the last glacial period.Line #136 The authors state "Statistical examination suggests that this δ 234 Ui excursion is restricted to..." with only a reference to Figs. 1 & 2. It is not clear what statistical tests are used and they should be specifically stated.Reply: We have divided the previous Fig. 2 into two separate figures: a revised Fig. 2 that exclusively displays the deep-sea coral δ 234 Ui data for different groups, and a new Fig.3, which features box plots for testing the statistical significance of large-scale trends and transient perturbations.In the updated Fig. 3, we employed a two-sided Wilcoxon rank sum test to assess the hypothesis of equal δ 234 U medians between this δ 234 Ui excursion and other δ 234 U data from This sentence is not clear to me.As written, it seems to suggest that the absence of high δ 234 U in the Pacific and Indian oceans during HS1 accounts for the δ 234 U spike in the Southern Ocean.Please either elaborate or rephrase for clarity.For clarity, reference Fig.2Cand state that the deep SS water does not show this.How is error from 230 Th correction calculated?Here the authors state that they use a 230 Th/ 232 Th of 5000.If used as a constant (i.e. with infinite precision), this should contribute no additional uncertainty to the calculation.However, in Supp.Fig 1, they show the effect of 232