The global carbon budget provides an important contribution to both science and policy by updating annually information about emissions and sinks of CO₂ and their trends. This year’s budget is no exception, with valuable insights including recent increasing trends in AGR in spite of a small increase in EFOS and that cannot be explained by the current models, and highlighting persisting uncertainty in SLand in the northern extra tropics. A notable new development is the inclusion of the RSS correction, which results in a weaker, but possibly more realistic, land sink. The report follows previous years’ structure and content, and is well written and well referenced. Nevertheless, I would like to point out a few aspects that can be improved or that need further clarification or discussion.

1) General comments

1.1) RSS correction

This is an important new addition to the budget, since the DGVM protocol used in previous budgets assumed no legacy emissions from land-use prior to 1850, implying a stronger land sink than what is plausible.

However, in many figures and tables, it is not clear whether the land component fluxes show include or not this correction, making it difficult for the reader to appreciate its impact in all relevant components of the budget.

1.2) Next-year projections

This budget, like previous ones, provides an estimate of the values of emissions and sinks for the next year, in this case 2025. These projections are based in very different methods, that rely on many assumptions and which are not consistent with each other (the ESM projections being the most self-consistent). Since next-year projections have been now given for several budgets, it would be important to discuss how much confidence can be attributed to these projections: how have previous projections performed? Are the respective uncertainty ranges provided here realistic?

2) Specific comments

Abstract

Lines 193-204: I understand the need to describe briefly the datasets used, but since the budget follows the same approach as previous years, I would recommend shortening this description, in order to rather highlight the new additions (for example the RSS correction) and provide more depth on the new insights from this year’s update (for example the very large value of BIM for 2024).

Lines 205-213: it would be helpful to report percent increase for all budget terms.

Lines 216-220: it would seem important to add here some insight about possible causes for the land BIM in 2024.

Introduction

Please state briefly in the introduction what the novel aspects of this year’s budget are (for example the RSS correction), compared to previous ones.

Lines 367-382: these sentences would probably fit better right in the beginning of the introduction, before stating the current budget’s approach.

Results

The use of sign to indicate an increase/decrease “by” is not always consistent but it is needed for clarity. For example, in Lines 996-1006 +3.6 is given for India, but – sign in “EU27 by” and “USA by” is missing. Many other instances can be found throughout the text across all sections, I note some lines here, but a careful editing is needed.  Lines: 1011, 1014-1015, 1028, 1039, 1042, 1048, 1055, 1057-61, 1078, 1205 (35TgC “up”), 1228, 1246, 1262, 1302, 1417, 1422, 1432, 1444, 1950-51.

Fossil fuel

Lines 1026: please state briefly what information and assumptions the projections are based on.

Land use change

Lines 1068-1069: clarify which period is covered by vegetation biomass observations.

Lines 1096-1099: it is the first instance when fluxes, rather than emissions are mentioned. It would be helpful for non-expert readers to state the sign convention here.

Lines 1103-1114: this is relevant for decision making, but it overlooks the fact that these models do not represent processes that can reduce the land sink, especially disturbance fluxes, as noted by Roebroek et al. (2023) among others. Please state limitations and assumptions underlying the values given here.

Lines 1111: it is unclear if here you refer to a potential re/afforestation sink, or current CDR.

Lines 1151: clarify here you mean nature-based CDR, right?

Lines 1160: are RECCAP2 the estimates that are shown in Figure 15?

Lines 1194-1196: I find this confusing, since BM models do not include fire emissions due to natural drivers such as El Niño droughts. Shouldn’t this be rather part of the SLand discussion? And do DGVMs capture these emissions?

Atmospheric CO₂

Lines 1271: is r for pearson correlation?

Lines 1295-1299: this is an important point. What would be the implications for 2023, then?

Lines 1301: briefly clarify what the “GCB regression method” is based on.

Lines 1303: give reference to “neutral ENSO year”.

Ocean sink

Lines 1365: remove . after Tropical Pacific

Lines 1382: is there a reason uncertainty ranges given using different notations (+-  vs [ ] )? If so, clarify.

Lines 1415-1420: it would be more informative to have this point being discussed along with discrepancies between models and observation-based products. Here makes comparison more difficult.

Lines 1443: briefly explain what predictors are used by the FFNN.

Lines 1489-1491: since there seems to be a linear relationship between the two, would it make sense to apply a correction to fCO₂? I do not have a strong opinion, but I am curious if the authors have considered this.

Land sink

Lines 1497-98: does the RSS correction affect only the mean values (bias), or are there carry-over effects that could result in different IAV and trends? Please clarify that the RSS correction was only applied to the DGVM multimodel mean, not to the individual DGVM estimates (as far as I understand).

Lines 1505-1506: give value of the residual sink here too.

Lines 1560: is it unclear if the values given here include the RSS correction, please clarify.

Lines 1571-1573: how about fires?

Lines 1574-1586: do the models representing fires estimate a weaker sink?

Partitioning of the fluxes

Lines 1683-1687:  it is worth mentioning that, however, the northern land flux does not seem to contribute much to the BIM in 2024.

Lines 1717-1721: are there implications for the BIM value in 2024 that can be drawn from this comparison?

Lines 1753-1795: this is a very informative addition. However, RECCAP2 provides estimates of regional budgets using additional datasets (possibly with higher confidence that global products). Therefore, it would be worth comparing the estimates here with the values provided by RECCAP2 for each region, or at least highlight those regions where results from RECCAP2 indicate large discrepancies between global datasets and refined regional products.

Closing the global carbon cycle

Lines 1873: th should be superscript

Lines 1876: with “historical” do you mean “pre-industrial”? Does this refer to the RSS correction?

Lines 1891-1893: is this only because of the RSS correction, or are there other reasons?

Lines 1894-1898: the discussion of the disagreements in AGR between the surface network and satellite-based GRESO should be linked here. What is the effect of these disagreements on BIM for 2024?

Lines 1910-1913: I understand a full attribution is out of scope in this already long report, but Bastos et al. (2021) showed that the BIM strongly correlated with differences between inversions and global models, so maybe something could be said about the BIM based on Figure 14? At first glance some of the peaks in BIM in Figure 4 seem to correspond to large differences between inversions and GCB estimates of the total and land fluxes shown in Figure 14, especially in the tropics.

Lines 2024-2032: for traceability, it would be useful to have a table reporting major corrections/updates to the data and their implications.

Lines 2057-2082: there are several satellite-based biomass products that now allow to derive at least some components of ELUC, for example Xu et al. (2021). Are these planned to be incorporated in future budgets, if not, what are the limitations?

Lines 2011-2116: provide values for the estimates by atmospheric inversions, oxygen-based estimates and Randerson et al. (2025) for comparison.  

Figures and Tables

Figure 13: do the DGVM estimates shown here include the RSS correction? Please clarify or, ideally, include both the uncorrected and corrected estimates.

Figure 14: does the gray shade for DGVMs correspond to the land-flux with or without RSS correction?

Figure 15 has very poor quality, please increase resolution.

Data availability

I agree that making the data available will contribute to greater understanding and new scientific insights of how the carbon cycle works, as stated in Section 7. However, in that spirit, why are gridded data provided openly for atmospheric inversions only, and not for the other datasets underlying the budget, especially those used in the main figures? Providing all gridded data openly would contribute to speed up understanding of uncertainties and knowledge gaps, by making it accessible to the broader scientific community. Furthermore, this hampers reproducibility of some of the main results of this paper (Figures 6, 11, 12, 15), and does not seem to fully align with ESSD core principles as stated in the journal’s page.

Supplement

It is difficult to extract from the supplementary information the key aspects of the protocol used for DGVM and BM models (starting year, period covered, forcing, adjustments, etc.), since they are referred to in between lengthy descriptions of specific aspects of the datasets. These lengthy descriptions are useful since they contain very relevant information, but I would suggest summarizing the modelling protocol and adjustments performed in a table (similar to Table S2 for GOBMs).    

Line 857: I suggest adding a subsection header for the RSS correction here (or highlight as a paragraph), making it easier for the readers to find the relevant information.

References cited

Caspar T. J. Roebroek et al. Releasing global forests from human management: How much more carbon could be stored? Science380, 749-753(2023). DOI:10.1126/science.add5878

 James T. Randerson et al.The weak land carbon sink hypothesis. Sci.Adv.11, eadr5489(2025). DOI:10.1126/sciadv.adr5489

Bastos, A et al. (2020), Sources of uncertainty in regional and global terrestrial CO2 exchange estimates. Global Biogeochemical Cycles, 34, e2019GB006393. https://doi.org/10.1029/2019GB006393

Liang Xu et al. Changes in global terrestrial live biomass over the 21st century.Sci. Adv.7,eabe9829(2021).DOI:10.1126/sciadv.abe9829

General:

This manuscript presents the 2025 update of the Global Carbon Budget as part of the established “living data” series. The paper maintains the high standards of transparency, community coordination, and documentation that characterise previous releases, and the resulting dataset will be of high value to both the research and policy makers.

Relative to the 2024 release, the manuscript introduces several methodological changes, most notably the RSS correction, the use of transient carbon densities in land-use change emissions, and the inclusion of satellite-derived atmospheric CO2 growth rates, and explicit data from Japan. These updates lead to noticeable changes in the land-ocean partitioning of CO2 uptake and in the behaviour of the budget imbalance. Overall, these changes are well documented in the Methods section.

The Results & Discussion highlights several key outcomes of the 2025 update, e.g. revised estimates of land and ocean sinks and updated uncertainty ranges. Satellite-based CO2 observations provide enhanced spatial and temporal coverage relative to surface networks. The 2024–2025 El Nino is highlighted as a major driver of elevated atmospheric CO2 growth rates and weakened sink performance. Emissions trends continue to diverge regionally, with declines in some industrialised regions and continued growth in developing economies. The remaining carbon budgets are updated, reinforcing the urgency of rapid emissions reductions. Remaining challenges include large uncertainties in land-use change emissions and ocean uptake, underscoring the need for expanded observations, improved regionalisation, and continued methodological development.

Overall the manscuript should be published of course. I nevertheless recommend some revisions primarily to improve clarity in distinguishing method-driven revisions from data-driven signals, and to tone down some interpretive statements in the Results and Discussion.

A central issue throughout the Results and Discussion is that several prominent changes relative to GCB 2024 arise primarily from methodological corrections (e.g. RSS corrections, revised ELUC treatment, updated GATM estimation), rather than from new observational constraints. While this distinction is carefully described in the Methods, it is not always carried through consistently into the Results and Discussion. I recommend that the authors more explicitly flag when differences relative to GCB2024 are driven by revised assumptions or corrections, especially when discussing the important differences relative to GCB2024 (particularly in ocean uptake, land–ocean partitioning, and budget imbalance)

The exceptionally large atmospheric CO2 growth rate in 2024, combined with a strongly negative budget imbalance, is a key result of this update. While the manuscript discusses this clearly, parts of the Discussion imply causal attribution (e.g. sink overestimation or sink failure) that is not uniquely supported by the budget closure alone. Given that multiple budget components (SLAND, SOCEAN, and GATM) were revised simultaneously, and that correlated uncertainties remain substantial, attribution of the imbalance to any single process is inherently ambiguous. Statements linking the 2024 imbalance to specific sink behaviour should therefore be softened or explicitly framed as hypotheses rather than conclusions.

Minor:

Please explicitly remind us readers that changes in ELUC relative to GCB2024 arise from revised model assumptions (transient carbon densities), rather than new land-use activity data.

When discussing the suspension of SOCAT-based products and applied corrections, a brief summary sentence reiterating the rationale would help non-specialist readers.

Minor inconsistencies remain in the use of terms such as “uncertainty,” “spread,” and “variability.” These could be harmonised for clarity.

Specific Comments (please be aware that the line numbers refer to the manuscript with tracked changes which I kindly received from the editors)

• Lines 641–642 (Abstract): The large negative budget imbalance in 2024 is interpreted as suggesting that “the total sink or GATM is strongly overestimated”. From a mass-balance perspective, an equally plausible interpretation is that land carbon losses are underestimated. Processes such as drought-enhanced respiration, disturbance-related emissions (e.g. fires, peat), and post-disturbance legacy fluxes are not fully constrained and would lead to an apparent negative imbalance. I suggest rephrasing (briefly in the abstract and in more detail in the main text) to explicitly acknowledge underestimated sources or land carbon losses as an alternative or additional explanation
• Line 1047: The sentence “Global emissions and their partitioning among the atmosphere, ocean and land are in balance in the real world” is potentially misleading. I suggest rephrasing 

• Lines 1222–1236: Please add a reference to the description of how aviation and shipping emissions are estimated and projected.
• Lines 1745–1747: Please add a reference to support the statement introduced here.
• Lines 2626–2627: Please provide quantitative values (absolute and/or relative) for the reported increase in CO2 emissions due to peat fires in tropical Asia.