Deglacial release of petrogenic and permafrost carbon from the Canadian Arctic impacting the carbon cycle

The changes in atmospheric pCO2 provide evidence for the release of large amounts of ancient carbon during the last deglaciation. However, the sources and mechanisms that contributed to this process remain unresolved. Here, we present evidence for substantial ancient terrestrial carbon remobilization in the Canadian Arctic following the Laurentide Ice Sheet retreat. Glacial-retreat-induced physical erosion of bedrock has mobilized petrogenic carbon, as revealed by sedimentary records of radiocarbon dates and thermal maturity of organic carbon from the Canadian Beaufort Sea. Additionally, coastal erosion during the meltwater pulses 1a and 1b has remobilized pre-aged carbon from permafrost. Assuming extensive petrogenic organic carbon oxidation during the glacial retreat, a model-based assessment suggests that the combined processes have contributed 12 ppm to the deglacial CO2 rise. Our findings suggest potentially positive climate feedback of ice-sheet retreat by accelerating terrestrial organic carbon remobilization and subsequent oxidation during the glacial-interglacial transition.

1. The authors hypothesize that oxidation of petrogenic OC released large amounts of CO2 to the atmosphere and thereby contributed to the rise in CO2 during the last deglaciation. This builds on the finding that mass accumulation rates of petrogenic OC to Beaufort shelf sediments were >10 times higher during deglacial warming periods than today, which suggests much higher petrogenic OC release from land at that time. Further, the authors assume that higher petrogenic OC release was accompanied by significant CO2 release. This assumption follows a hypothesis that is currently under debate, e.g. in the non-peer-reviewed preprint in ref. 13, which states that petrogenic OC oxidation may have contributed to the deglacial rise in CO2. However, this concept is highly uncertain. Indeed, it seems that there is scattered evidence that glacial excavation of bedrock in glaciated areas may release petrogenic OC and locally cause higher CO2 emissions than in non-glaciated areas (e.g., Horan et al., 2017). Yet, the exposure of fresh bedrock and possibly petrogenic OC is outbalanced by the drawdown of atmospheric CO2 through silicate weathering, a process that is known to contribute to long-term removal of atmospheric CO2 and modulating the Earth climate system over glacial-interglacial time scales. In addition to chemical weathering, changes to in the biosphere and inland permafrost may have profound impact on the OC balance throughout the last deglaciation. This is even specifically stated for the Mackenzie drainage by the literature cited in the paper (Horan et al., 2019;ref 43 in the manuscript). Unfortunately, these processes are not well considered or appropriately discussed in the present version of the paper. Furthermore, the budget calculations for the cumulative release of 84 Pg as CO2 during the last deglaciation is insufficiently described. Given the lack of observational evidence and in the light of the published literature on this topic, I'm skeptical about the author's findings about significant release of CO2 from petrogenic OC oxidation.
2. The authors leverage dual-isotope source apportionment (13C and 14C) to distinguish between past release of terrOC from petrogenic sources and permafrost. As end member for petrogenic OC, the mixing model includes published 13C ratios of source rocks within the Mackenzie river catchment and assumes 14C-free OC. As end member of the terrestrial biosphere, 13C ratios of soils and pre-depositional 14C ages of HMW nalkanoic acids are used. There are a number of issues with these end members I would like to address: i) The chosen end members do not account for terrOC release from older permafrost deposits from the non-glaciated parts or OC-rich glacial till in Arctic Canada, while such deposits are present in this area, e.g. on Hershel Island and along the Yukon coast, and contribute to significant terrOC release.
ii) These end members cannot distinguish between any terrOC from soils or other permafrost deposits that were buried below the Laurentide Ice Sheet. Would this be another possible source of terrOC that may have contributed CO2 during the last deglaciation?
Further, the compound-specific 14C age as end member for the terrestrial biosphere is problematic for several reasons: iii) HMW n-alkanoic acids may not solely represent "fresh" biospheric terrOC but is most likely a mixture of terrOC integrated from several sources over this large area. To my knowledge, it may also not be ruled out that these compounds partly derive from petrogenic OC directly, or are affected and contaminated by microbial recycling of petrogenic OC. iv) HMW n-alkanoic acids represent only a small fraction of the terrestrial organic matter but is here used to represent terrOC at the bulk level. HMW n-alkanoic acids are, however, comparatively degradation-resistant and may survive longer in the environment than others, which may create a bias towards higher 14C ages. For the revised version of this paper, the authors should consider an improved end member selection for the source apportionment of the OC.

Minor points
Line 25: As stated above, I see only limited support for this statement.
Line 54-57: The studies mentioned (ref. 10 and 16) provide only evidence for petrogenic OC release. I see no support for "substantial OCpetro mobilization and oxidation…".
Line 66-67 "more work is required to test this hypothesis": That's also my understanding.
Introduction: The introduction needs to explain how CO2 is produced from petrogenic OC. Is it only a process of chemical weathering or is also microbial decomposition involved?
Line 135: How is burial efficiency defined in this study? Does is represent i) the preservation of OC between release on land and long-term burial in marine sediments or ii) the loss of OC between sediment deposition at the sediment-water interface and residual OC buried during early diagenesis processes? Line 155: Consider a different word choice than catastrophic here and elsewhere in the paper.
Line 161: How old were the oldest HMW-FA? This needs to be stated to follow the meaning of this sentence.
Line 167: How is the composition expected to be different that it would result in a different end member age?
Line 176-178: The "delay" of an event may also result from lateral transport of sediment OC. Could this process affect the dating of this or other events in this study?
Line 179 "Although the distinct increases in TOC MAR were linked…": add "likely" or similar. I suggest to use more careful phrasing here and elsewhere in the paper.
Line 180: What parameters of bulk OC? Looking at the 14C of bulk OC I see very large changes (i.e. a drop of the pre-depositional age from ~30 to ~20 kyr). How do you explain this difference if its not a change in source or composition?
Line 213 and 219-222: Wouldn't also the CO2 drawdown by chemical weathering increase? As described above, I think this needs to be included in this discussion to reflect on the net CO2 exchange with the atmosphere.
Line 234-236: We can or we may expect? This and other sentences of this paragraph are rather speculative and needs to be more carefully phrased.
Line 236: Sediments reactive surface areas are also considered to sequester OC to form mineral-organic associations, which stabilize and protect OC from remineralization. I would argue that this process rather limits the reactivity of the OC.
Line 243-252: What area (km2) was used here? Are glaciated areas considered to produce CO2 as well? Was the area allowed to change through time with regard to glacier retreat? More background information is needed to follow this calculation.
Line 256: What is the flux estimate for 10-yr old shales is based on? This number is 20 times higher than the measured/estimated CO2 flux in this region. If this is only assumed I think an alternative scenario needs to be presented, in which the age of the substrate has no effect to the oxidation rate/CO2 production. This may also replace the uncertainty calculations in line 257-259. Wu et al. investigated two cores collected from the Beaufort Sea to study the carbon dynamics since the last deglacial time. They concluded that there is a significant amount of CO2 released through the oxidation of petrogenic carbon and the erosion of permafrost. They tried to build on the previous theory framework and made quantitative calculations to demonstrate the importance of these two processes in explaining the variation of atmospheric CO2 concentrations and C isotopes. Overall, the authors presented attractive and robust data echoing those published previously, and this study further confirms the theory of terrestrial OC oxidation contributing to deglacial atmospheric CO2 fluctuations. Here, I am providing some concerns, questions, and suggestions that may help improve the MS. Major questions: 1) Petrogenic carbon's oxidation and permafrost OC's erosion have previously been proposed as possible mechanisms substituting the ocean ventilating theory in explaining deglacial atmospheric CO2 fluctuations. This study tried to make sense quantitatively; while reading through the MS, I doubted the calculation, relatedness, and contribution from this study. First of all, the estimate of the amount of OCpetro bases heavily on an equation published in a Biogeoscience Discussion, which is still not accepted by Journal Biogeosciences. As a critical parameter of this equation, the oxidation rate (0.89) was adopted from another study (Horan et al., 2019). However, while reading through this key reference, the only place that mentioned this same oxidation rate was under the discussion of carbonate weathering. Additionally, the calculation of weathering rate is not constant, as deglacial retreat would largely thicken the soil and thus decrease the weathering rate. Overall, I doubt the calculated 84.3PgC release through the OCpetro erosion. Furthermore, I do not see the linkage between the budget calculation and the core measurements from this study. This study is on the sedimentary record from the Beaufort Sea, while the budget calculation of OCpetro oxidation is purely based on hinterland data published previously. Second, this study further reinforced the importance of permafrost erosion as an essential source of atmospheric CO2 during the deglacial time. However, I do not see how much this study contributed rather than repeating what was published by . Overall, I doubt the carbon budget calculation and the contribution solely from this study.
2) It is always important to consider water depth variation while interpreting data generated from a sediment core offshore. Since the last glacial time, the sea level has increased by ~170m. During the deglacial time, the location where sediment cores were retrieved was much shallower. What matters more is that the sampling locations are closer to the coast. Such deglacial-interglacial differences could largely explain the sedimentation rate differences. Explicitly, the sedimentation rate is expected to be much higher when the sampling location was closer to the coastline during the deglacial time. It essentially explains higher OCpetro and OCpermafrost MARs at the deglacial time. Additionally, the results and discussion rely heavily on one sediment core, while the other core is only presented with limited data and discussion. '" 4<9 9A8$@9@69D @B89? 5A8 F<9 5EEB7=5F98 D9EG?FE 8BG6F98 @9% -=DEF B: 5??# F<9 M&', was adopted from Hilton (Fig. 4b); hence, the error generated from the three endmember-model could be enormous. 4) The application of some "classic" parameters. Rock-Eval is commonly applied as a method in estimating the thermal maturity of source rocks. However, Rock-Eval analysis on modern biomass, soil, and sediments generates similar Tmax values. The observed variation of Tmax downcore could fully be explained by the diagenetic alteration over the past 14kyr and is not a robust proxy in demonstrating petrogenic carbon inputs. +9E=89E# :OO =E ABF H5?=8 I<9A JBG 5EEG@9 F<9 C9FDB;9A=7 75D6BA =E @5FGD9 BD BH9D$ mature already since thermal cracking breaks down organic molecules. Additionally, the EB=? <5E 699A :BGA8 FB CDB8G79 E=;A=:=75AF 5@BGAF B: OL <BC9A9E# D5F<9D F<5A OO hopenes, even when the soil is "fresh". Other comments: Line 57: The above evidence only supports mobilization, but not oxidation of OCpetro. Lines 87-89: thermal maturity based on Rock-Eval, in this study, does not indicate petrogenic OC inputs. Lines 204: deglacial does not cause physical erosion; glacier advancing does. Lines 204-209: shorter transport distance is more reasonable than the intense physical erosion of bedrock in explaining enhanced OCpetro MAR. Lines 207-209: The burial efficiency might be higher during the deglacial time since these particles have been deposited along floodplains and have gone through preliminary oxidation. Lines 219-240: this section presented a contradictory discussion on whether OCpetro is extensively protected or weathered. The first part argues that OCpetro is protected due to mineral protection, while the second part claims that the OCpetro is heavily weathered due to high surface area. Lines 239-258: The discussion of the budget is based on several criteria, which essentially is not related to the core of this study, which is what generated from the sediment cores. Lines 241: Blattmann (2021) was not accepted and should not serve as an essential reference here. Additionally, I do not see this equation being presented in Blattmann (2021, biogeoscience discussion). Line 251: This is not the number from the original study (Horan et al., 2019). Lines 259-273: the discussion of permafrost relies heavily on ; hence it is hard to extract the new contribution from this study. Line 268-270: residual permafrost/soil on the floodplain or what has been deposited at the coastal region could supply aged permafrost carbon to the study area. This permafrost has been exposed to the atmosphere or has gone through fluvial transport and hence is more resistant to degradation.
Reviewer #3 (Remarks to the Author): Reviewer #3 attachment on the following page.

Dear Wu et al.,
The manuscript presents a record of organic matter burial near the mouth of the Mackenzie River covering the episode of most recent deglaciation towards the present day. The authors characterize the organic mater using radiocarbon and biomarkers. From this, a timeline of changing sources and burial rates of permafrost, OCpetro, marine, and terrestrial biospheric OC is reconstructed. From the burial rates, OCpetro oxidation rates in the catchment are estimated and together with published data a quantitative estimate on the combined contribution of permafrost and OCpetro derived CO2 is generated which is placed into a global context.
The work presents a novel dataset from a remote location that is strategically chosen to provide best possible insight into the dynamics of deglaciation. The authors have chosen to interpret their dataset in a completely new way with respect to the dynamic behavior of OCpetro both in terms of its reburial flux and its decay along its source-to-sink trajectory. These are basic constraints that are conspicuously missing and just beginning to emerge in the literature (e.g., Berg et al., 2021). For the first time, the study of Wu et al. provides us with a view of changes taking place across an entire cycle of deglaciation in an area recording the history of the largest continental ice sheet retreat on Earth.
In my opinion, the current manuscript is in need of major structural improvement and refinement of arguments and technical precision. By addressing the comments elaborated below, the current work can greatly increase its persuasiveness and potential to energize the community to investigate the dynamic role of terrestrial stocks of permafrost and petrogenic OC on atmospheric chemistry and climate across glacial-interglacial cycles. This work is of great value for providing new perspectives and basic quantitative constraints on the carbon fluxes that have governed the fate of our planet in our geologically most recent history.

Zurich
Major comments and suggestions: The Results section and Discussion section should either be strictly separated or merged. Currently, considerable discussion resides in the results part (e.g., 132-141, and many other bits and pieces) and leads to some repeats in the manuscript. In conjunction with this, the permafrost vs. OCpetro components of the interpretation are presented in a confusing manner and I got the impression that the authors were unsure about their own interpretation. Regarding the presentation of the results, if the authors decide to merge their results and discussion, I suggest that the authors use the bulk radiocarbon data to build their core story and use other datasets (e.g., dated compounds, GDGTs, etc.) as complementary components to build the discussion. The bulk radiocarbon data seem to provide a clear interpretation around which the complementary datasets provide interesting discussion avenues.
The definition of terrestrial biosphere and distinguishing or equating that with permafrost carbon is unclear. For example, in lines 173-174 the HMW-FAs were interpreted as stemming from permafrost and immature OCpetro. However, in the source apportionment in lines 452-454, HMW-FAs ages are used to represent the age of terrestrial biospheric organic carbon. So is terrestrial biospheric organic carbon equated with permafrost carbon? And in this case, is OCpetro also considered terrestrial biospheric carbon of ancient origin? lead to some confusion. I recommend the authors refine the usage of their terminology to make the interpretation clearer.
How was the Monte Carlo simulation implemented? How was the uncertainty of the marine reservoir age considered? Seems like two ages were considered, but how was this implemented numerically? Was the uncertainty of the age model of the core also incorporated in the calculation of uncertainty for the different pools of OC? Given the uncertainty of the age model at the base of the core(s), this seems warranted.
What is the uncertainty/reproducibility of the Rock-Eval results for Tmax? For a non-expert, the approximately 12°C spread in Tmax (Fig. 2g) seems narrow and a discussion of the uncertainty appears warranted.
Please include (representative) chromatograms for fatty acids, alkanes, etc. for documentation purposes in the supplementary information. Additionally, I felt like the visual display of total mass accumulation rates and total organic carbon content was missing for interested readers to get a feel for the data. I think these data would reside well in the supplementary information. Furthermore, I imagine that a graphical display of mass accumulation rate of the different types of TOC with the resolution made possible by the source partitioning presented in Fig. 4 would also be a valuable addition to the supplemental and could be helpful in aiding the discussion.
The quantitative estimates of carbon release from permafrost appears to solely come an Asian-based study  and OCpetro oxidation presented largely hinges on assuming a constant burial efficiency and assuming that the sedimentary records are representative for the integrated catchment (sediment dispersal or sorting seems unaddressed). The restricted occurrence of permafrost in the Mackenzie Basin is mentioned (e.g., lines 171, 273), so is the permafrost-released CO2 from this area considered negligible? Additionally, OCpetro mass accumulation rates from the site investigated is claimed to represent deglaciated North America (lines 310-311), which misses inputs from the Alaskan fjords and Canadian west coast  and from the catchments draining into the Hudson Bay, which also drain OCpetro-rich areas of the Western Canadian Sedimentary Basin, as well as moraines (line 241) and glacial lakes (Blattmann et al., 2019) which trap OCpetro in these intermediate sinks. Skeptical readers may have a hard time with the patchwork carbon budget and buying into the idea of the constant burial efficiency, but I think there is an opportunity to present this as a conservative estimate.

Minor comments:
The terms remobilization and oxidation are sometimes used imprecisely. For example, in line 27 remobilization is used. However, remobilization per se is a carbon neutral process and in its context a positive climate feedback is proposed. There are a few such instances in the manuscript. Please refine the usage of these terms.
There are several points of discussion surrounding the mode of permafrost mobilization/decay, including coastal erosion, shelf flooding, and hinterland thawing. However, from a carbon budget perspective, it is unclear why distinguishing between these different modes of permafrost mobilization and decay is important. I think adding reasons or motivation behind answering these questions would be valuable for the readers.
Line 36: Anomaly of the atmosphere or ocean? Please specify.
Line 135: In this context, how relevant is isostatic uplift over this timescale?
Line 147: What kind of fieldwork evidence is this referring to? Please specify. ectation that may or may not agree with reality. Lines 298-304: This sentence needs to be fixed.
Line 452: Concerning the notation: I suggest using separate notation for age-corrected 14 C concentrations, as I assume you are using age-corrected 14 C values for assessing the end members?
Lines 131, 473: Reference 32 seems like a weak as this study looked at a specific time in the Neoproterozoic. Large swaths of the catchment are Cretaceous and Devonian (see geological map of Canada). Recovering data hidden in other studies seems worthwhile to get a representative picture on the stable carbon isotope composition of OCpetro, or discussing the representativity of the available constraints.

Language comments:
The singular and plural forms for bedrock are the same: bedrock. See dictionaries Merriam Webster, etc.
Please refine text for smoother reading. For example: Line 63: Please review the usage of articles (the, a, an) and possibly have a native speaker proofread.

Author responses to reviewer comments on the manuscript entitled "Deglacial release of petrogenic and permafrost carbon from the Canadian Arctic impacting the carbon cycle" (NCOMMS-21-36247) by J. Wu and co-authors
We thank all three reviewers for the efforts spent on reviewing and we greatly appreciate the constructive and helpful comments. All comments were carefully considered and most of them were incorporated into the revised version of the manuscript. We believe the current version has been significantly improved based on these comments.
V^fZ]^ln[lmZgmbZe k^oblbhgl Z\\hk]bg`mh ma^k^ob^p^kl fZchk \hg\^kgl, pab\a Zk^: 1. All reviewers have major concerns about our mixing model, endmembers, and endmember values. We thus compiled a Supplementary Text to discuss in detail how we define endmembers and endmember values (updated) and how we test the robustness of mixing model outcomes. The outcome is also updated in the Main Text. 2. All reviewers raise concerns about our budget calculation and have questions about the linkage between the budget calculation and the measurements made on the sediment samples. We think this is because of an insufficient description of our approach. Therefore, we made major revisions to more clearly explain the concept of our approach. 3. Reviewer#1 points out weaknesses in our argument, particularly considering the scattered evidence of higher OCpetro oxidation in glaciated regions. In the revised version, we re-structured our argument and emphasize that our intention is to explore the potential magnitude of OCpetro release based on the emerging evidence rather than derive hard numbers. Besides, a key part missing in the initial version is the discussion of other weathering processes as well as OCterr-bio burial. We thus added discussions of these processes to the Main Text. 4. Reviewer#2 points out the limitations of some classic parameters to specifically indicate thermal maturity. We carefully changed our interpretation of their variations and used them more as supporting evidence. 5. We added graphical displays in the Supplementary Information, e.g., chromatograms, total MARs, and the MARs of different types of carbon.
Below we present our point-by-ihbgm k^lihgl^mh ma^k^ob^p^kl \hff^gml.

Sa^Znmahkl arihma^lbs^maZm hqb]Zmbhg h_ i^mkh`^gb\ NB k^e^Zl^] eZk`^Zfhngml h_ BN2
to the atmosphere and thereby contributed to the rise in CO2 during the last deglaciation. This builds on the finding that mass accumulation rates of petrogenic OC to Beaufort shelf sediments were >10 times higher during deglacial warming periods than today, which suggests much higher petrogenic OC release from land at that time. Further, the authors assume that higher petrogenic OC release was accompanied by significant CO2 release. This assumption follows a hypothesis that is currently under debate, e.g. in the non-peer-reviewed preprint in ref. 13, which states that petrogenic OC oxidation may have contributed to the deglacial rise in CO2. However, this concept is highly uncertain. Indeed, it seems that there is scattered evidence that glacial excavation of bedrock in glaciated areas may release petrogenic OC and locally cause higher CO2 emissions than in non-glaciated areas (e.g., Horan et al., 2017). Yet, the exposure of fresh bedrock and possibly petrogenic OC is outbalanced by the drawdown of atmospheric CO2 through silicate weathering, a process that is known to contribute to longterm removal of atmospheric CO2 and modulating the Earth climate system over glacialinterglacial time scales. In addition to chemical weathering, changes to in the biosphere and inland permafrost may have profound impact on the OC balance throughout the last deglaciation. This is even specifically stated for the Mackenzie drainage by the literature cited in the paper (Horan et al., 2019; ref 43 in the manuscript). Unfortunately, these processes are not well considered or appropriately discussed in the present version of the paper. Furthermore, the budget calculations for the cumulative release of 84 Pg as CO2 during the last deglaciation is insufficiently described. Given the lack of observational evidence and in the light of the publila^] ebm^kZmnk^hg mabl mhib\, Hf l\^imb\Ze Z[hnm ma^Znmahkl _bg]bg`l Z[hnm lb`gb_b\Zgm release of CO2 from petrogenic OC oxidation.

Reply 1.1:
We thank the reviewer for these insightful comments, pointing out the rather limited database on the process of OCpetro oxidation following glacial excavation, and highlighting the counter-acting process of silicate weathering. We would like to stress, though, that we do by no means ignore the effects of silicate weathering, but that we attempt an estimate of the potential impact the commonly not considered process of petrogenic OC oxidation might have.
During the LGM, the bedrock was overlain by the Laurentide Ice Sheet and Cordilleran Ice Sheet which were up to a thousand-meter thick. The complete retreat of ice sheets must have eroded and exposed the long-term isolated OCpetro to the atmosphere. We observed significant OCpetro mobilization, while whether it was largely oxidized is unclear due to the currently incomplete understanding of this process. Unfortunately, we are not able to study such extreme processes in the contemporary Earth system or in the next few centuries. Therefore, here we demonstrate our attempt to explore an alternative scenario of large CO2 emissions in response to enhanced erosion, based on evidence from past periods of rapid warming and glacial melting. We argue that the OCpetro flux (10.7±6.6 tC km -2 yr -1 ) we estimated for the last deglaciation is within a reasonable range compared with the compilation of Blattmann (2022) 1 according to which OCpetro oxidation fluxes from rock-disseminated forms of kerogen under aerobic conditions amount to 0.3-64 MgC/km 2 /yr and the fluxes from catchments with ongoing deglaciation range from 9-50 MgC/km 2 /yr 2 . In the Main Text, we re-structured our argument and clarified that our study makes an attempt at exploring a new scenario. Please see lines 212-223.
We agree that multiple weathering processes and OCterr-bio burial must have been involved and contributed to the net CO2 budget. Our study, based on organic carbon dynamics, focuses on the climate impact of the single process of OCpetro oxidation, while the discussion of the net carbon budget is missing in our initial version. We added more discussions on these processes and the net CO2 budget during the glacial period. Our study may contribute to the discussions on OC-related processes, while a comprehensive understanding of the role of glacial retreat or a net carbon budget needs more studies on both inorganic and organic processes. Please see our discussion in lines 343-370.
Besides, we would like to point out that the reference of Blattmann (2022)

1.2
The authors leverage dual-isotope source apportionment (13C and 14C) to distinguish between past release of terrOC from petrogenic sources and permafrost. As end member for petrogenic OC, the mixing model includes published 13C ratios of source rocks within the Mackenzie river catchment and assumes 14C-free OC. As end member of the terrestrial biosphere, 13C ratios of soils and pre-depositional 14C ages of HMW n-alkanoic acids are used.
There are a number of issues with these end members I would like to address: Reply: As Reviewers #1 and #2 have major concerns about our mixing model, we here explain our ideas before replying to each comment. In general, there are two ways to define endmembers in a mixing model. One is based on the type of terrOC, i.e., petrogenic or biospheric (e.g., ref 3,4 ). Notably, in such a definition, terrOC is apportioned regardless of the place of origin. Another approach is based on the spatial distribution of terrOC. For example, Grotheer et al. (2020) 5 define terrOC endmembers in the Beaufort region by attributing carbon to the Mackenzie River catchment or Herschel Island. However, in such a definition, a complex mix of organic matter including petrogenic OC is characterized in each system and we cannot estimate the relative contributions of different types of organic material independently. Both ways of definitions can be used for a mixing model but both have limitations. Defining endmembers based on their spatial distributions usually cannot cover all possible sources, whereas when defining endmembers by their origins it is usually difficult to distinguish OCpetro and old OCbioshphere.
There are two key reasons why we believe it is better here to define endmembers according to their origin. First, the key finding of this study is the enhanced petrogenic OC input during the last deglaciation, therefore our aim is to estimate its contributions. Second, the molecular-level radiocarbon dating (F 14 CFA) can represent all possible OCterr-bio sources as a whole and help to distinguish between OCpetro and old OCterr-bio.
i) The chosen end members do not account for terrOC release from older permafrost deposits from the non-glaciated parts or OC-rich glacial till in Arctic Canada, while such deposits are present in this area, e.g. on Hershel Island and along the Yukon coast, and contribute to significant terrOC release.

Reply 1.2-i:
We respectfully disagree with this comment and would like to elaborate on our approach. Our assumption is based on the fact that functionalized compounds like fatty acids are considered to be absent in mature organic materials like petrogenic carbon that has undergone diagenesis and partly even katagenesis. On the other hand, long-chain n-alkyl compounds are widely used biomarkers for terrestrial higher plants, and their compoundspecific age can reflect combined contributions to sediment from a variety of thermally immature sources including vegetation, soils, permafrost, or even older deposits like peats or lignite 6•8 . Therefore, the compound-specific radiocarbon ages we obtained on long-chain fatty acids are taken to reflect the mean 14C content of the complex mixture of all materials rich in terrestrial organic matter contributing to our sediments, including old permafrost deposits or OC-rich glacial till.
Further, we suspect that Reviewer#1 suggests us to include 13C values from Herschel Island and Yukon coast when we define the endmember values for terrestrial biospheric carbon. In the revised version, we include 13C values from the Herschel Island retrogressive thaw slumps (n=7) and onshore samples from the Yukon Coastal Plain (n=19) 9 . Overall, the mean 13C value is still within the range defined in our initial version. Please see Supplementary Information for more details.
ii) These end members cannot distinguish between any terrOC from soils or other permafrost deposits that were buried below the Laurentide Ice Sheet. Would this be another possible source of terrOC that may have contributed CO2 during the last deglaciation?
Reply 1.2-i: As elaborated above, our study aims at distinguishing different types of organic matter and not different regional contributions. The hypothetical OC-rich permafrost deposits buried below the Laurentide Ice Sheet could be a possible carbon source once the glacial retreat has exposed it to the atmosphere, and we acknowledge that we cannot distinguish its contribution from permafrost that did not reside below a glacier.
We did not evaluate the significance of such hypothetical subglacial organic-rich permafrost relicts due to the following reasons. First, in most of the glaciated regions, organic-rich permafrost deposits did not form, and only the regions that remained unglaciated during any of the glacial stages might form such organic-rich permafrost. According to the ice-sheet configuration described in Batchelor et al. (2019) 10 , such areas in North America are much smaller than in other regions and mainly existed in the south and east of North America. Our core location is not suitable to document such changes. Considering that such organic-rich permafrost deposits are typically unconsolidated, we estimate their potential to be covered by glaciers rather than being eroded completely during the glacial advance to be rather unlikely. Second, our records show relatively small biospheric OC fractions during the last deglaciation (even during the rapid coastal erosion events), suggesting an overall smaller contribution of permafrost carbon at least in Northern North America. Due to these reasons, our material might not be ideal to address the importance of permafrost carbon in this region.
Further, the compound-specific 14C age as end member for the terrestrial biosphere is problematic for several reasons: iii) HMW n-ZedZghb\ Z\b]l fZr ghm lhe^er k^ik^l^gm _k^la [bhlia^kb\ m^kkNB [nm bl fhlm likely a mixture of terrOC integrated from several sources over this large area. To my knowledge, it may also not be ruled out that these compounds partly derive from petrogenic OC directly, or are affected and contaminated by microbial recycling of petrogenic OC.
iv) HMW n-alkanoic acids represent only a small fraction of the terrestrial organic matter but is here used to represent terrOC at the bulk level. HMW n-alkanoic acids are, however, comparatively degradation-resistant and may survive longer in the environment than others, which may create a bias towards higher 14C ages.

Reply 1.2-iii and 1.2-iv:
We agree that there are shortcomings of HMW n-alkanoic acids to represent the terrestrial biospheric carbon. As mentioned in comments 1.2-iii and 1.2-iv, the HMW n-alkanoic acids may contain (small) petrogenic contributions and only represent a small fraction of terrOC which may be comparatively degradation-resistant. These shortcomings may cause a bias towards higher 14C ages. Despite that, we here explain the reasons and advantages of using HMW n-alkanoic acids as endmembers as follows: (1) In the study of Drenzek et al. (2007) 4 (from the Beaufort Sea), it has been found maZm Hg this case, however, the depleted isotopic compositions of the pyrolysis products suggest that long-chain fatty acids and alkanes are likely representatives h_ m^kk^lmkbZe NB hg ma^pahe^•.
(2) Importantly, the vegetation, permafrost, and ice sheets in North America were dynamic during the last deglaciation. The relatively fixed endmember values from the contemporary system cannot reflect such changes and thus might not be applicable to the paleo system. The HMW n-alkanoic acids, as a whole to represent all OCterr-bio sources (e.g., freshly produced OC from plants, OCterr-bio from permafrost deposits, and OCterr-bio from variable soil profiles in nonpermafrost regions), are expected to reflect the dynamic changes in terrestrial biospheric carbon composition over time, although we acknowledge that one single compound group does not represent the full diversity of organic matter present in such a system.
(3) As mentioned above, HMW n-alkanoic acids represent all possible sources as a whole, relieving the constraint to find proper endmember values to represent all possible sources.
(4) Endmember values determined on bulk OM on land are fixed while the radiocarbon signals of terrestrial OC may change (e.g., via degradation and re-suspension) during transport to the core location. This may cause an underestimate of the OCterr-bio contribution. In contrast, using HMW n-alkanoic acids values determined in the study material circumvents these possible additional complications.

Please see Supplementary Information for discussions on defining OCterr-bio endmember values.
For the revised version of this paper, the authors should consider an improved end member selection for the source apportionment of the OC.
In the Supplementary Information, we updated the definitions or added discussions on 13C values for OCterr-bio and OCpetro (explained below) endmembers. As for F 14 C values for OCterr-bio endmember, as mentioned above, we believe the signals in HMW n-alkanoic acids have obvious advantages over published bulk-level endmember values.

Minor comments
Line 25: As stated above, I see only limited support for this statement.
We agree with this comment and rephrased it. Please see lines 26-27.
Line 54-57: The studies mentioned (ref. 10  Introduction: The introduction needs to explain how CO2 is produced from petrogenic OC. Is it only a process of chemical weathering or is also microbial decomposition involved?
We thank for this comment and we add in the Introduction an explanation of the CO2 release from OCpetro. Both chemical weathering and microbial decomposition may be involved. Please see lines 53-56.
Line 135: How is burial efficiency defined in this study? Does is represent i) the preservation of OC between release on land and long-term burial in marine sediments or ii) the loss of OC between sediment deposition at the sediment-water interface and residual OC buried during early diagenesis processes?

Sa^[nkbZe^__b\b^g\r bg mabl lmn]r bg]b\Zm^l ma^ik^l^koZmbhg h_ NB [^mp^^g k^e^Zl^hg eZg]
and long-term burial in marine sediments. We now clarified it in the manuscript and please see lines 141-142.
Line 155: Consider a different word choice than catastrophic here and elsewhere in the paper.
Line 161: How old were the oldest HMW-FA? This needs to be stated to follow the meaning of this sentence.
Done. The pre-depositional ages of HMW-FA have been indicated in lines 167-168.
Line 167: How is the composition expected to be different that it would result in a different end member age?
Thanks for pointing this out. We suggest a different OCterr-bio composition from a perspective of the organic carbon age. In our study, the OCterr-bio is defined as a mixture of various sources. It ]h^l ghm hger k^ik^l^gm _k^la OC but also pre-aged OCterr-bio that could be derived from permafrost deposits or deeper soil profiles, or even highly-degraded OCterr-bio. During the last deglaciation, the LIS has restricted vegetation development, resulting in fewer contributions from young/fresh OC. Therefore, we propose that the mean age of OCterr-bio should be much older than that of the contemporary system. This is why we chose to use HMW-FA age as endmember values instead of a fixed endmember value from the contemporary system. We added an explanation for this. Please see lines 170-173.  13 , and the radiocarbon dates of these two cores are consistent during the last deglaciation. If we include the AMS14C dates from the lower part of core JPC15/37, the first event in our core may have an older age. Therefore, we believe the delay is kZma^k Zg ZiiZk^gm ]^eZr Zg] \Zg be attributed to a less-constrained chronology or that our archive has only documented part of the event. We added discussions. Please see lines 185-188.

Kbg^179 @emahn`a ma^]blmbg\m bg\k^Zl^l bg SNB L@Q p^k^ebgd^]~: Z]] ebd^er hk lbfbeZk.
I suggest to use more careful phrasing here and elsewhere in the paper.
Done. Please see, for example, line 189.
Line 180: What parameters of bulk OC? Looking at the 14C of bulk OC I see very large changes (i.e. a drop of the pre-depositional age from ~30 to ~20 kyr). How do you explain this difference if its not a change in source or composition?
The decrease in pre-depositional age from 30 to 20 kyrs is probably attributed to vegetation development since the LIS retreated. Vegetation development may result in a younger mean age of OCterr-bio (ie^Zl^, \hfiZk^mh Q^ier Kbg^167), [nm bm ]h^l ghm g^\^llZkber f^Zg Z eZk`^k contribution from OCterr-bio to the sedimentary OC. The younger OCterr-bio (kb\a bg • 14 C) may cause a younger apparent age of bulk OC. This is also supported by our mixing model output which shows decreasing bulk OC ages while OCpetro dominated during the last deglaciation (Fig 4).
Line 213 and 219-222: Vhne]gm Zelh ma^BN2 ]kZp]hpg [r \a^fb\Ze p^Zma^kbg`bg\k^Zl^? As described above, I think this needs to be included in this discussion to reflect on the net CO2 exchange with the atmosphere.
The CO2 drawdown by silicate weathering may also increase during the deglaciation. In the manuscript, we focus more on the OC dynamics and its climate impact. At the end of the manuscript, we added discussions about other processes and emphasized the importance of a comprehensive understanding of the net CO2 budget. Please see lines 343-370.
Line 234-236: We can or we may expect? This and other sentences of this paragraph are rather speculative and needs to be more carefully phrased.
Done. The sentences are carefully phrased. Please see lines 251-259.
Line 236: Sediments reactive surface areas are also considered to sequester OC to form mineral-organic associations, which stabilize and protect OC from remineralization. I would argue that this process rather limits the reactivity of the OC.
As pointed out by Reviewer#3, OCpetro in many cases probably already comes associated with lithogenic minerals. Therefore, our previous argument that mineral protection may cause higher burial efficiency of OCpetro during the last deglaciation is not correct. We deleted this argument.
The effect of organic matter protection on mineral surfaces that the reviewer mentioned here is likely more important for fresh biospheric organic carbon, like freshly produced phytoplankton debris that attaches to the mineral surface is protected from oxidation through this adsorption (or its entrainment in pores).
Line 243-252: What area (km2) was used here? Are glaciated areas considered to produce CO2 as well? Was the area allowed to change through time with regard to glacier retreat? More background information is needed to follow this calculation.
We thank the reviewer for this comment and we improved our description of the approach. We assume that outcropping shales that were previously covered by ice sheets would start oxidation upon exposure to the atmosphere. Therefore, the Aexposure indicates the areas with outcropping shales that were freshly exposed during the ice-sheet retreat. V^^fiaZlbs^_k^laer^qihl^] Zk^Z• [^\Znl^p^have considered dynamic changes in both the exposed area and the oxidation flux over time. We assume that the freshly exposed shales generated a higher oxidation flux and this flux decreases over time. Therefore, the Aexposure is calculated based on the shales distribution from Amiotte Suchet et al. (2003) 14 and the changes in ice-sheet extent from Peltier et al. (2015) 15 and is calculated for every 500 years to indicate the freshly exposed areas in different time periods. The results are shown in Figure 5a. Please see the revised text in lines 260-273.
Areas that remained glaciated are not considered to produce CO2, only when the areas with shale outcrops were exposed to the atmosphere do we start to calculate CO2 release from these areas.
Line 256: What is the flux estimate for 10-yr old shales is based on? This number is 20 times higher than the measured/estimated CO2 flux in this region. If this is only assumed I think an alternative scenario needs to be presented, in which the age of the substrate has no effect to the oxidation rate/CO2 production. This may also replace the uncertainty calculations in line 257-259.
The flux estimate is based on an assumption that the weathering rate of OCpetro decreases with substrate aging, following the equation Foxidation = F0 × t -0.7116,17 . F0 is the oxidation flux of freshly produced substrate (10-yr old), and Foxidation denotes the oxidation flux of substrate at age t. To calculate F0, we need a known Foxidation with a known substrate age t. Here we use the oxidation flux of 1 tC km -2 yr -1 which is based on the estimate from Horan et al. (2019) 18 (0.89±0.32 tC km -2 yr -1 ) in the modern shales-dominated region. This region was exposed between 10-15 cal. kyr BP 19 , and thus we assume a substrate age of 10±5 kyrs. We are then able to estimate the flux for 10-yr old shales (F0) based on the modern oxidation flux and assumed substrate age. We revised the text to make it clearer. Please see lines 273-293.
Line 262: Is this 84 Pg budget based on entire North America or only the Mackenzie river catchment? This does not become clear in this paragraph or elsewhere in the manuscript.
We assume the OCpetro oxidation flux from the shales-dominated regions in the Mackenzie River catchment is applicable to entire North America, and thus the budget is calculated for entire North America. We added more information and please see lines 268-295.  20 . Our modeling exercise is updated relative to the published results in that a new version of the carbon cycle model was used, and that the age of permafrost carbon is set to 10 kyr, which is more realistic in light of our new findings. More importantly, in our exercise, we consider the combined effects of permafrost carbon mobilization and degradation and oxidation of petrogenic OC.
Reply to Reviewer #2: 2.1 Petrogenic carbon's oxidation and permafrost OC's erosion have previously been proposed as possible mechanisms substituting the ocean ventilating theory in explaining deglacial atmospheric CO2 fluctuations. This study tried to make sense quantitatively; while reading through the MS, I doubted the calculation, relatedness, and contribution from this study. First of all, the estimate of the amount of OCpetro bases heavily on an equation published in a Biogeoscience Discussion, which is still not accepted by Journal Biogeosciences. As a critical parameter of this equation, the oxidation rate (0.89) was adopted from another study (Horan et al., 2019). However, while reading through this key reference, the only place that mentioned this same oxidation rate was under the discussion of carbonate weathering. Additionally, the calculation of weathering rate is not constant, as deglacial retreat would largely thicken the soil and thus decrease the weathering rate. Overall, I doubt the calculated 84.3PgC release through the OCpetro erosion. Furthermore, I do not see the linkage between the budget calculation and the core measurements from this study. This study is on the sedimentary record from the Beaufort Sea, while the budget calculation of OCpetro oxidation is purely based on hinterland data published previously. Second, this study further reinforced the importance of permafrost erosion as an essential source of atmospheric CO2 during the deglacial time. However, I do not see how much this study contributed rather than repeating what was published by . Overall, I doubt the carbon budget calculation and the contribution solely from this study.

Reply 2.1:
We thank the reviewer for this comment. Overall, we believe that ma^k^ob^p^kl major concerns about our budget calculation and the relatedness are due to an insufficient description of our approach in the initial version. We made substantial revisions to this part. Please see lines 260-296. Below we reply to each question in this comment.
The equation (Jcarbon = Aexposure × Texposure × Foxidation) for budget calculation is valid because the unit of oxidation flux (tC km -2 yr -1 ) indicates that it requires a multiplication of the area by the time over which the oxidation occurs to calculate the amount of carbon released. Therefore, we deleted the citation of Blattmann (2022), although we also note that the manuscript is now published as a peer-reviewed paper. We have not clearly described our concept of how we account for decreasing oxidation fluxes with increasing exposure time in the initial version. In the revised version, we explained that the oxidation flux is known to be 0.89±0.32 tC km -2 yr -1 in the contemporary system 18 and is estimated to be 10.7±6.6 tC km -2 yr -1 during the last deglaciation. The changes in oxidation fluxes may reflect processes such as soil formation and vegetation development. To include such processes in our calculation, we assume that the oxidation flux decreases with increasing exposure time (Foxidation = F0 × t -0.71 ). Please see our revisions in lines 273-289.
The linkage between budget calculation and core measurements is not clearly described in our initial version. The concept is that today, the shales-dominated regions in the Mackenzie River catchment have a modern OCpetro oxidation flux of 0.89±0.32 tC km -2 yr -118 . Due to the increase in OCpetro MAR (12±6 times) in marine sediments and an assumption of constant burial efficiency, we postulate that the past oxidation flux in the Mackenzie River catchment also increases 12±6 times (i.e., 10.7±6.6 tC km -2 yr -1 ). Thereby, the linkage is established between the increased OCpetro MAR and the past oxidation flux on land. We here assume that OCpetro oxidation fluxes in the shales-dominated regions from the Mackenzie River catchment can be applied to other shales-dominated regions in North America. Furthermore, based on the modern oxidation flux 18 and the estimated past oxidation flux, we simulate the long-term behavior of OCpetro oxidation flux that decreases with increasing exposure time, following the equation Foxidation = F0 × t -0.71 . F0 is the oxidation flux of freshly exposed substrate (10-year old), and Foxidation denotes the oxidation flux of the substrate at age t. Simulation of the long-term behavior requires a F0 and please see the reply above to Line 256• for how F0 is derived. Although the long-term behavior of Foxidation is implemented based on the modern oxidation flux, we compared with our estimated flux that falls well within the equation and stands for a substrate age of <100 to 2000 years (indicated in Fig. 5b), thus to some degree supporting our assumption. Please see our revisions in lines 226-230, 236-238, 274-275, and 290-293.
First of all, our carbon cycle simulation includes both OCpetro release during the glacial retreat and permafrost carbon release from coastal erosion, as both processes may have released terr-OC that is related to the ice-sheet retreat. In addition, as explained in the manuscript, previous studies on permafrost carbon remobilization were carried out in largely unglaciated regions and it is difficult to unambiguously determine the process of coastal erosion. It means that the carbon release proposed by   20 is tentative. Our study confirms that coastal erosion during the rapid sea-level rise was a major process to remobilize permafrost carbon. This finding corroborates the scenario of pulsed carbon release as found in   20 and convinces us to include this process in our simulation as it is related to icesheet retreat. Moreover, we used an updated model scenario compared to   20 , including an adjustment of the assumed age of eroded permafrost deposits (changed to 10 kyrs), reflecting the growing database.

2.2
It is always important to consider water depth variation while interpreting data generated from a sediment core offshore. Since the last glacial time, the sea level has increased by ~170m. During the deglacial time, the location where sediment cores were retrieved was much shallower. What matters more is that the sampling locations are closer to the coast. Such deglacial-interglacial differences could largely explain the sedimentation rate differences. Explicitly, the sedimentation rate is expected to be much higher when the sampling location was closer to the coastline during the deglacial time. It essentially explains higher OCpetro and OCpermafrost MARs at the deglacial time. Additionally, the results and discussion rely heavily on one sediment core, while the other core is only presented with limited data and discussion.

Reply 2.2:
We thank the reviewer for this comment and agree that the sea level may influence the distance between the coast and core location, which further influences the sedimentation rate. However, the Beaufort Sea has a much narrower shelf than other Arctic Seas (Fig. 1a). As shown in Figure 1b, the red area indicates the flooded shelf since the Last Glacial Maximum. Based on a rough estimate, the coastline retreat (since the LGM) has increased the distance between the river mouth and core location by ~110 km, which is much less than in the extensively flooded Eurasian continental shelves. The number may be further reduced if we only consider the changes during the last 14 kyrs. We added an argument for higher MARs due to a shallower water depth during the last deglaciation. Please see lines 247-248.
This study focuses more on the core ARA04C/37. In our previous study, we have carefully compared the chronology and magnetic susceptibility between core ARA04C/37 and core JPC15   13 . The two cores have similar chronology and magnetic susceptibility and documented the same timing of the YD flood. More importantly, there are common bits of black particulate matter found in both cores, which are likely petrogenic. We believe the strong OCpetro input was the most important feature of deglacial terrestrial OC remobilization and both cores have documented that. Because the study we have carried out for core ARA04C/37 is of a heavy workload, we thus only analyzed samples of B/A interval from core JPC to extend our records and further support the scenario of strong OCpetro input.

2.3
molecular-level signature circumvents the constraint of unknown processes during transport which may change the radiocarbon signals from land to the core location. Therefore, using F14CFA has obvious advantages over its limitations.

2.4
Sa^Ziieb\Zmbhg h_ lhf^\eZllb\ iZkZf^m^kl. Qh\d-Eval is commonly applied as a method in estimating the thermal maturity of source rocks. However, Rock-Eval analysis on modern biomass, soil, and sediments generates similar Tmax values. The observed variation of Tmax downcore could fully be explained by the diagenetic alteration over the past 14kyr and is not a kh[nlm ikhqr bg ]^fhglmkZmbg`i^mkh`^gb\ \Zk[hg bginml. A^lb]^l, _ bl ghm oZeb] pa^g rhn assume the petrogenic carbon is mature or over-mature already since thermal cracking breaks down organic molecules. Additionally, the soil has been found to produce significant amount h_ ahi^g^l, kZma^k maZg ahi^g^l,^o^g pa^g ma^lhbe bl _k^la.

Reply 2.4:
We gratefully thank for this comment. We may have overinterpreted the variations in the initial version and we agree that there are limitations of these classic parameters. As proposed by many studies 4,23•25 , the contemporary Mackenzie River system has significant contributions from OCpetro and all our parameters are in agreement with this characteristic and do not conflict with each other. Besides, the bulk OC ages and the mixing model outcome all suggest an enhanced OCpetro input during the last deglaciation. We, therefore, believe that the ab`a SfZq, ehp BOH, Zg] ehp _ during the last deglaciation are indicative of large OCpetro input. In the revised manuscript, we deleted the overinterpretation of data variations and used these classic parameters as supporting evidence for our findings. Please see our revisions in lines 122-133.

Minor comments
Line 57: The above evidence only supports mobilization, but not oxidation of OCpetro.
Done. Please compare to the reply to Reviewer#1 (reply to lines 54-57). We rephrased it and please see lines 65-66.
Lines 87-89: thermal maturity based on Rock-Eval, in this study, does not indicate petrogenic OC inputs.
Done. We clarified that we use these parameters as supporting evidence. Please see lines 96-97.
Lines 204: deglacial does not cause physical erosion; glacier advancing does.
The sentence is deleted in the revised version.
Lines 204-209: shorter transport distance is more reasonable than the intense physical erosion of bedrock in explaining enhanced OCpetro MAR.
We agree this may be an important factor causing enhanced OCpetro MARs. However, the ice extent in Dalton et al. (2020) 19 shows a largely unglaciated Mackenzie River basin since 12 cal. kyr BP (please see figure below), while our source apportionment still indicates the dominance of OCpetro input between 12-8 cal. kyr BP. This, we think, is most likely caused by intense physical erosion of bedrock. Lines 207-209: The burial efficiency might be higher during the deglacial time since these particles have been deposited along floodplains and have gone through preliminary oxidation.
Here we defined the burial efficiency as the preservation of OC between release on land and long-term burial in marine sediments. Hence, such a process, e.g., preliminary oxidation along floodplains would not increase the burial efficiency under our definition. We added the definition of burial efficiency in lines 141-142.
Lines 219-240: this section presented a contradictory discussion on whether OCpetro is extensively protected or weathered. The first part argues that OCpetro is protected due to mineral protection, while the second part claims that the OCpetro is heavily weathered due to high surface area.
Thanks for pointing this out. In our initial version, we tended to discuss on the molecular level the effect of organic matter protection on the mineral surface. However, this may not be the case for OCpetro, we, therefore, deleted this argument. Please compare to reply to Reviewer#1 (reply to line 236).
Instead, we argue for increased OCpetro oxidation due to the high surface area. Rock fragments that already contain organic matter as an intrinsic part of the entire matrix will allow more access (of oxygen or microbes or both) to the organic matter the larger the surface area is relative to the mass. So smaller ground rock fragments offer more access to organic matter than larger fragments.
Lines 239-258: The discussion of the budget is based on several criteria, which essentially is not related to the core of this study, which is what generated from the sediment cores.
Please compare our reply to comment 2.1.

3.4
What is the uncertainty/reproducibility of the Rock-Eval results for Tmax? For a non-expert, the approximately 12°C spread in Tmax (Fig. 2g) seems narrow and a discussion of the uncertainty appears warranted.

Reply 3.4:
We thank for this comment and agree that we may have overinterpreted the Tmax variations of approximately 12°C in the initial version, which is relatively narrow. In the revised version, we used Tmax and other parameters to support the evidence for overall strong OCpetro input in this region, while avoiding further interpretation of their variations. In this study, evaluating changes in OCpetro contribution is achieved by using the dual carbon isotope mixing model. Please compare the reply to 2.4 and see our revisions in lines 96-97 and 122-133.

3.5
Please include (representative) chromatograms for fatty acids, alkanes, etc. for documentation purposes in the supplementary information. Additionally, I felt like the visual display of total mass accumulation rates and total organic carbon content was missing for interested readers to get a feel for the data. I think these data would reside well in the supplementary information. Furthermore, I imagine that a graphical display of mass accumulation rate of the different types of TOC with the resolution made possible by the source partitioning presented in Fig. 4 would also be a valuable addition to the supplemental and could be helpful in aiding the discussion.

Reply 3.5:
We thank for this comment. We have included representative chromatograms for FAs and alkanes in Supplementary Figures 7-8. The TOC content and total mass accumulation rates of core ARA04C/37 have already been published in Wu et al. (2020) 13 . However, since we have included two more AMS 14 C dates in this study, we showed updated total MARs in Supplementary Figure 6.
We also included the graphical displays of MARs of different types of organic carbon in Supplementary Figure 5. Due to the three strong events (i.e., the YD flood event and two more coastal erosion events), defining the TOC MARs that are used to calculate different OC MARs requires caution. We discussed three different ways to define TOC MARs between 14-10 cal. kyr BP (when the three events occurred). The TOC MARs that we used to calculate OCpetro MARs in the Main Text is a conservative estimate. Please see more information in the Supplementary Discussion.

3.6
The quantitative estimates of carbon release from permafrost appears to solely come an Asian-based study  and OCpetro oxidation presented largely hinges on assuming a constant burial efficiency and assuming that the sedimentary records are representative for the integrated catchment (sediment dispersal or sorting seems unaddressed). The restricted occurrence of permafrost in the Mackenzie Basin is mentioned (e.g., lines 171, 273), so is the permafrost-released CO2 from this area considered negligible? Additionally, OCpetro mass accumulation rates from the site investigated is claimed to represent deglaciated North America (lines 310-311), which misses inputs from the Alaskan fjords and Canadian west coast  and from the catchments draining into the Hudson Bay, which also drain OCpetro-rich areas of the Western Canadian Sedimentary Basin, as well as moraines (line 241) and glacial lakes (Blattmann et al., 2019) which trap OCpetro in these intermediate sinks.
Skeptical readers may have a hard time with the patchwork carbon budget and buying into the idea of the constant burial efficiency, but I think there is an opportunity to present this as a conservative estimate.

Reply 3.6:
We thank the reviewer for pointing out that the way we present the quantitative estimate of carbon release may not be sufficiently clear. It is correct that we refer to a previous study based on a core from the East Asian continental margin when estimating carbon released from permafrost. However, in the study by   20 , the assumption is made that the core record is representative of a large-scale process active across the entire area that lost most of its permafrost cover during the last deglaciation, namely sea-level rise induced mobilization of permafrost deposits that had accumulated on areas that were exposed during the glacial and were flooded during the deglaciation. Making this assumption,   20 use published estimates of the carbon contained in these flooded areas (105°E-128°W; please see figure below), as well as estimates of the fraction of this organic matter that is remineralized, to arrive at a quantitative estimate of the effect that permafrost mobilization might have had on atmospheric CO2. This assumption seems to be justified, as several records published after this study reported similar periods of increased permafrost carbon mobilization, and we also find comparable deglacial OC MAR maxima that we relate to sea-level rise.
In our study, we take a similar approach to derive a quantitative estimate of C release from exhumed petrogenic carbon. We postulate that our core record is representative of the process of petrogenic carbon mobilization in response to the deglaciation of North America. The variation in OCpetro flux is thus taken to reflect erosion/exposure of OCpetro, much of which might remain on land and could be oxidized following erosion. We then continue by estimating the amount of petrogenic carbon that was exposed and might have been eroded in the regions experiencing glacial ice retreat based on reconstructions of ice margins through the deglaciation combined with geological maps of outcropping shales in these regions (Figure 5a). We then, using modern observations on OCpetro oxidation fluxes and assumptions on decreasing oxidation rates with increasing exposure time (Figure 5b), estimate OCpetro oxidation flux across the exposed area (Figure 5c). Note that for this approach we took to estimate oxidation fluxes, we consider mainly processes occurring on land and take our core record simply as an indicator of the temporal increase in OCpetro exhumation on land. It is therefore irrelevant by which pathway the (likely rather small) fraction of exhumed OCpetro that accumulated in the ocean was transported. We thus do not need to consider fluxes to Alaskan fjords or the

Minor comments
The terms remobilization and oxidation are sometimes used imprecisely. For example, in line 27 remobilization is used. However, remobilization per se is a carbon neutral process and in its context a positive climate feedback is proposed. There are a few such instances in the manuscript. Please refine the usage of these terms.
Done. Please see line 29-30. There are several points of discussion surrounding the mode of permafrost mobilization/decay, including coastal erosion, shelf flooding, and hinterland thawing. However, from a carbon budget perspective, it is unclear why distinguishing between these different modes of permafrost mobilization and decay is important. I think adding reasons or motivation behind answering these questions would be valuable for the readers.
Line 36: Anomaly of the atmosphere or ocean? Please specify.
Done. Please see line 39.
Line 135: In this context, how relevant is isostatic uplift over this timescale?