the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 31 Jul 2026
The Global Ocean Data Analysis Project version 3 (GLODAPv3) – an internally consistent biogeochemical data product for the world ocean
Abstract. The Global Ocean Data Analysis Project (GLODAP) is a synthesis effort providing surface-to-bottom ocean biogeochemical observations determined through chemical analysis of discrete bottle samples, with an emphasis on seawater inorganic carbon chemistry and related variables. Version 3 of GLODAP comprises data from 1181 cruises, spanning more than 50 years of observations (1972–2023). It includes all data from the previous GLODAPv2.2023 (Lauvset et al., 2024) together with newly added data from 57 cruises. For all cruises, 13 core variables (temperature, salinity, oxygen, nitrate, silicate, phosphate, dissolved inorganic carbon, total alkalinity, CFC-11, CFC-12, CFC-113, CCl₄, and SF6) have undergone extensive quality control with particular focus on the identification and removal of systematic differences between cruises. The data are available in two formats: (i) as submitted by the data originators, converted to World Ocean Circulation Experiment (WOCE) exchange format, and (ii) as a merged data product in which adjustments have been applied. These adjustments were determined using crossover analyses in combination with a newly developed global inversion method, the furthest-first routine. The applied adjustments are intended to remove systematic differences arising from differences in measurement methods, calibration, and/or data-handling practices, while preserving known or likely temporal trends and natural variability. The consistency of the adjusted data product is estimated to be 0.0013 for salinity, 0.7 % for oxygen, 0.4 % for nitrate, 0.5 % for silicate, 0.5 % for phosphate, 1.2 µmol kg⁻¹ for dissolved inorganic carbon, and 1.4 µmol kg⁻¹ for total alkalinity. Consistency estimates could not be derived for transient tracers, but they are believed to be consistent to better than 5 % (10 % for SF₆). The enhanced consistency enables different datasets to be used together with greater confidence. Newly introduced cruise-specific uncertainty estimates for all core variables provide more granular quantifications of remaining cruise-to-cruise inconsistencies. Additional variables, including pH, discrete CO₂ fugacity (fCO₂), isotopic tracers, and others, were not subjected to secondary quality control but are included in the data product.
The original data, their documentation (metadata), and DOIs are available through the Ocean Carbon and Acidification Data System (OCADS) of NOAA’s National Centers for Environmental Information (NCEI), which also hosts the merged data product. All secondary quality control decisions and supporting information can be found in the online adjustment table (https://glodapv3.geomar.de, last accesses 26.06.2026). The product is distributed as a single global file and as four regional subsets (Arctic, Atlantic, Indian, and Pacific Oceans) under https://doi.org/10.25921/m6tp-mj50 (Lange et al., 2026). These adjusted files also include ancillary and approximated data obtained through interpolation or calculation from measured data.
Competing interests: At least one of the (co-)authors is a member of the editorial board of Earth System Science Data.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(3731 KB) - Metadata XML
-
Supplement
(183 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2026-496', Anonymous Referee #1, 26 Aug 2026
-
AC1: 'Reply on RC1', Nico Lange, 15 Oct 2026
Reviewer comments: normal font
Author replies: bold font
General Comment:
GLODAPv3 is the third full re-evaluation of the synthesis, with this time 1181 cruise files, about 1.50 million samples, 1972–2023. Relative to v2.2023 it replaces the WLSQ inversion with the furthest-first (FF) algorithm, lets the regional clusters emerge from the crossover network rather than prescribing them, reduce the minimum adjustment limits (DIC and Talk) by a factor 2, adds per-cruise uncertainties, and adds an explicit trend-preservation step. All these advances are clear. In particular the FF inversion is a striking improvement. Section 4.2 is very insightful, no previous GLODAP release has stated that clearly that the network contains no direct Atlantic-Pacific connection, or that the Indo-Pacific link rests on one cruise. My comments are mostly about the text, not the analysis. The one place where I think the paper could be clearer and more thorough in its explanation is the carbonate system: the choice of input pair, the choice of constants, at a moment when the literature on constants has moved considerably, how users should treat pH data. Nothing below requires new data. Overall, congratulations and thank you on behalf of the community to the authors for putting this together.
Thank you for your positive assessment of the manuscript and constructive feedback. See specific comments below for content related responses.
Specific Comments:
The consistency numbers in the abstract are not the quantity that the earlier versions reported. The abstract values (1.2 µmol kg-1 DIC, 1.4 TAlk …) are the adjusted CWMO wRMSE of Table 9. In the v2 family the quoted consistency was the minimum adjustment limit. Lauvset et al. (2024) say that those numbers are "not uncertainties but rather a priori estimates of global inter-cruise consistency". So going from 4 to 1.2 µmol kg-1 can be read as a threefold improvement in data quality by the reader, but that is not the case. This should somehow be made clearer. It also sits awkwardly with Eq. 2. With Amin = 2 µmol kg-1 and σ about 0.8, the product ships ucruise ≈ 2.15 µmol kg⁻¹ for DIC while the abstract advertises 1.2. I guess my confusion here could be fixed with a few sentences, reporting ucruise so that the number is comparable with v2 and with the product files, or keeping the wRMSE and better explain what it is and what the v2 equivalent was. The abstract should also make clear that these values describe consistency within regional clusters and that inter-basin consistency is not constrained by the present network.
We see that the consistency values quoted in the abstract, i.e., the wRMSE of the CWMOs, could be interpreted as being directly comparable to adjustment limits (reported for earlier GLODAP releases or used for this version), which is not the case. Neither should these values be interpreted as indicating a threefold improvement in data quality or be confused with cruise-specific uncertainty estimates. In particular, the cruise-specific uncertainty estimates incorporate the minimum adjustment limits used to represent the limits of confidence in the adjustment procedure and are therefore not directly comparable to the wRMSE of the remaining cruise inconsistencies. However, the wRMSE of the CWMOs is the metric that directly describes consistency at the data-product level. Note that a closely related metric was also reported in previous releases (Table 7 in Lauvset et al., 2024, defined as ‘inter-cruise consistency’, and Table 5 in Olsen et al., 2016, defined as ‘global consistency’).
We have revised the manuscript to make the distinction between these different metrics clearer (lines 598–605) and made the link between the wRMSE and remaining cruise inconsistencies explicit in the abstract. Further, we added to the abstract that these estimates represent consistency within regional clusters, as robust conclusions regarding global consistency cannot be made given the characteristics of the present network. Lastly, we added Figure S3 to the supplementary material showing an overview of the uncertainty estimates.
The numbers in section 3.2 and in Table 8 disagree for most core variables. Table 8 is internally perfect, every column sums to 1224, and unadjusted + adjusted + (-888) + (-777) reproduces "with data" exactly, but numbers in the text don’t agree (e.g. the text says there are 998 entries with TAlk data, table says 851). Alkalinity needs a separate look because the paragraph does not add up on its own terms (998 - 30 - 282 = 686, not 544), and the gap to Table 8 is larger than for other variables.
The numbers in the text were unfortunately not updated after the final revision of the product. Table 8 contains the correct numbers, whereas the corresponding numbers in Section 3.2 were outdated. We have now updated the text to be consistent with Table 8. The alkalinity paragraph has also been corrected so that the numbers add up.
The pH data are left in an ambiguous state for users. The reasoning in section 4.1 is careful and I agree with the decision. But looking at it from an outside user perspective. GLODAPv2.2023 shipped adjusted pH with a stated consistency of 0.01 to 0.02. GLODAPv3 ships unadjusted pH, no consistency estimate, and nine cruises that earlier versions had excluded are back in. The message for user is confusing, can we still trust and use pH data the same way it was possible to do in the earlier versions. Please develop the text around this, explain how to pH estimates, what uncertainty to assume, whether measured pH or pH calculated from adjusted TCO2-TAlk is preferable and under what conditions, whether v2-adjusted and v3-unadjusted pH can be combined in a time series.
We agree that the change from adjusted pH in GLODAPv2 to unadjusted pH in GLODAPv3 requires clearer guidance for users. We have therefore expanded Sections 2.2.7, 4.1, and 5.3. Specifically, in Section 4.1, we now explicitly state that the absence of secondary QC in GLODAPv3, as well as the re-inclusion of data, should not be interpreted as a reduction in pHT data quality relative to GLODAPv2. Rather, by avoiding potentially inappropriate adjustments given the limitations of a universal pHT secondary QC (as discussed in the section), GLODAPv3 leaves users able to, and places the responsibility on them to, assess whether individual pHT data, or subsets thereof, are fit for their particular application. Further, in Section 5.3, we now explicitly state that uncertainty should be evaluated on a case-by-case basis and provide guidance on how to do so depending on the application. In these pH-specific recommendations, we also address the use of measured versus calculated pHT. While not explicitly added to the manuscript, we hope that the current text in combination with these additions makes clear that GLODAPv2 and GLODAPv3 data should not be combined. Note that the pHT adjustment limits applied in previous GLODAP releases (0.01–0.02, depending on the update) were based on a priori estimates of the level of inter-cruise consistency that could reasonably be expected in the data product. While we still consider 0.02 to be a reasonable upper threshold for identifying data that may warrant further investigation, there have been no objective methods for achieving or verifying such a level of universal consistency. These values should therefore not be interpreted as formally estimated cruise-level uncertainties or as data-product consistency estimates, either for GLODAPv3 or for previous releases.
The carbonate-system choices could be more justified. Section 2.2.7 states that "for calculations of TAlk or TCO2, only pHT data were used, and fCO₂ data were not considered", with no reason given. Section 4.1 says that there are inconsistencies beyond uncertainty propagation when fCO₂ is combined with TCO2 or TAlk (García-Ibáñez et al., 2022) but that concerns calculated pH, not the use of fCO₂ as an input for TAlk or TCO2 (and the reference is missing from the list). "Currently no consensus exists on which set of constants should be used (Carter et al., 2024b)" is weaker than the literature cited in the preceding sentence. Fong and Dickson (2019), Woosley (2021), Woosley and Moon (2023), Schockman and Byrne (2021), Sulpis et al. (2020) all point out to disagreement in the Lueker et al. constants and provide updated sets of constants. Lueker may remain the best compromise for a global legacy product, but that case should be made rather than deferred to an absent consensus. I wonder given the temperature and salinity range over which the chosen set of constants is validated, what fraction of GLODAPv3 falls outside.
We have revised Sect. 2.2.7 to clarify that the constants and input variables used in the GLODAPv3 calculations were selected primarily to ensure consistency with previous GLODAP releases. In particular, fCO₂ was not used as an input for the calculation of TAlk or TCO₂ because the primary purpose of these calculations is gap filling consistent with the legacy GLODAP approach. Also, please note that the constants of Schockman and Byrne (2021) and Sulpis et al. (2020) have limited temperature ranges and wouldn’t be applicable globally. Further, Woosley and Moon (2023) do not provide new constants, but found that each set of constants does fine under some conditions and poorly under others. Moreover, we now explicitly state that GLODAP is not intended to provide an authoritative carbonate-system product. Together, this provides the rationale for retaining the same set of constants, input variables, and output variables as in previous GLODAP releases. We have also clarified that the use of the selected equilibrium constants should not be interpreted as implying that they are universally preferred. Alternative sets of equilibrium constants have been proposed in the literature, and these may be more appropriate for specific applications. Sect. 5.3 further notes that “[…] recalculation of carbonate system variables using alternative equilibrium constants or a different combination of input variables, for example including fCO₂, may be preferable.” Lastly, we now mention that extra care should be taken when using calculated carbonate-system variables for which the temperature and salinity conditions fall outside the validated range of the chosen set of constants. This applies to approximately 25% of the calculated carbonate-system data (T: 2–35 °C; S: 19–43).
Technical Corrections:
Lines 227. Did you mean PyCO2SYS, rather than CO2SYSv2?
The incorrect citation was used in the previous draft. We used the MATLAB implementation of CO2SYS, integrated into our Python pipeline using GNU Octave. We have updated the reference accordingly, replacing Humphreys et al. (2022) with van Heuven et al. (2011).Table 5 footnote a still lists pH among adjusted variables.
Footnote updated accordingly.Table 9 labels salinity in ppm, I am confused if that is a typo or if I am not following something.
That was a typo and is now corrected for.Citation: https://doi.org/10.5194/essd-2026-496-AC1
-
AC1: 'Reply on RC1', Nico Lange, 15 Oct 2026
-
RC2: 'Comment on essd-2026-496', Enhui Liao, 12 Sep 2026
This manuscript presents GLODAPv3, a valuable contribution to the oceanographic and broader Earth science communities. The authors have gone beyond data collection by developing and applying methods to evaluate consistency among cruises, identify systematic offsets, preserve temporal trends, and provide additional uncertainty information. These methodological developments strengthen the scientific utility of the product and support the combined use of observations from different periods, regions, and measurement programs. Overall, I consider this an important dataset and a worthwhile contribution to ESSD. My comments primarily concern discrepancies in the reported numbers, consistency between the text and figures or tables, and clarification of several methodological descriptions. Addressing these points would improve the clarity, reproducibility, and usability of the manuscript.
- Lines 600–601 (p. 25) state that salinity data from 152 entries required adjustment, whereas Table 8 (p. 23) reports 158 adjusted entries. The other salinity counts agree, including the 861 entries remaining after exclusions. Please reconcile the adjustment count and ensure that the associated figures use the same final selection. Similar discrepancies occur in the counts reported for other variables; please check all numerical summaries against the final dataset.
- The seven action categories in Table 4 (p. 10) sum to 1225 for both salinity and oxygen. However, lines 126–127 report 1181 cruise files, and lines 549–551 report 1224 QC entries. Table 4 describes its units as “cruises,” although “entries” may be intended. Please clarify the statistical unit, and use “cruises” and “entries” consistently throughout the manuscript.
- Lines 48–50 and 126–127 state that GLODAPv3 contains all GLODAPv2.2023 data plus 57 newly added cruises, yielding 1181 cruises. The cited GLODAPv2.2023 publication reports 1108 cruises; 1108 + 57 = 1165, leaving a difference of 16 cruises. Please reconcile these totals by specifying any additions, reinstatements, removals, or changes in cruise definitions.
- Lines 625–644 (p. 25) distinguish pre-2010, 2010–2018, and post-2018 Japanese silicate data. Lines 895–896 (p. 32), however, refer to preadjustments to pre-2010 and post-2019 Japanese data. These descriptions treat 2019 differently. Please specify the intended year intervals and use them consistently.
- Equation (1) (p. 17) is printed as a sum of Wr/W. This does not express the weighted mean described in lines 385–398 because the denominator lacks the sum of the weights. Please write the numerator as Σj(Wijrij) and the denominator as ΣjWij, and define the summation set. Consider using a symbol such as r̄i for CWMO, rather than x̄, since x already denotes the adjustment vector in lines 344–346.
- Lines 349–350 state that shorter time separations receive higher weights. Lines 353–355 state that the additional factors are normalized so that low values indicate high confidence, multiplied together, rescaled to 1–2, and then multiplied by the inverse variance. If the final rescaling preserves the ordering, less confident crossovers would receive larger weights. Please provide the complete scaling equations, including the normalization bounds and whether the final mapping reverses the ordering.
- Lines 347–348 define the base weight as the squared inverse of the offset standard deviation. The Figure 5 caption (lines 359–361) instead describes the inverse standard deviation as the weight. Please reconcile the caption and methods with the implemented calculation.
- Lines 381–382 state that a trend is retained if α exceeds 1.96 times its standard error. Please clarify whether the intended criterion is |α| > 1.96 SE(α), so that significant negative trends are also retained.
- In Figure 10 (p. 23), the legend defines the light-blue, orange, and red categories as <2 × limit, 2–4 × limit, and >4 × limit. The caption (lines 575–577) instead defines them as below the minimum, between one and two times the limit, and above twice the limit. These definitions differ by a factor of two. Please reconcile the legend and caption. Please also specify the year range represented by “2010s”.
- The Figure 6 caption (p. 16) states that the figure shows three main processes, whereas the figure labels steps 1–4 and the text also describes four steps. Please make it consistent.
- The Table 7 caption (lines 438–440) states that qualifying cruises are adjusted in the third step. Lines 406–415 define step 3 as selection of the “Cruises-to-Adjust” list, while lines 417–419 place the final adjustment inversion in step 4. Please reconcile these descriptions.
- Table 10 (p. 33) labels the phosphate row “PO2,” whereas the manuscript otherwise uses PO4. Please correct the variable label.
- Table 5 and its footnote (p. 12) include pH among the variables receiving additive adjustments. Lines 222–225 and 739–740 explicitly state that secondary QC was not applied to pH in GLODAPv3. Please remove pH from the applicable adjustment list.
- Table 9 labels salinity as ppm and reports 1.9 → 1.3, whereas lines 604–605 report 0.0019 → 0.0013 for the same wRMSE. Please check the unit.
- The reference to Müller and Gruber, “Progression of ocean interior acidification over the industrial era,” is dated 2034 at lines 1287–1288. The paper was published in 2024. Please correct the year.
- Line 227 cites “CO2SYSv2 (Humphreys et al., 2022),” but the corresponding reference at lines 1204–1205 is “PyCO2SYS v1.8: marine carbonate system calculations in Python.” Please verify the software implementation and version intended here and correct the citation accordingly.
- Gruber et al. (2019), cited at line 106, has no matching entry in the reference list. Other citations also appear to be missing. Please cross-check all in-text citations against the reference list and add missing entries or correct the author–year citations as appropriate.
- A challenge is that crossover differences contain both measurement biases and real ocean variability. One possible option would be to sample existing multiyear or multidecadal ocean biogeochemical model output at the actual cruise locations, dates, and depths. A case without imposed measurement biases could test whether the method introduces unnecessary corrections to real ocean changes. Cases with imposed cruise biases and calibration shifts could test its ability to recover known errors. Such experiments could quantify false adjustments, residual errors, regional trend distortion, and performance in weakly connected parts of the crossover network. Where appropriate, they could also test whether corrections inferred from deep water remain suitable for the upper water column. This suggestion does not need to be implemented in the current revision, but it could be discussed in the manuscript or considered for future work.
- Line 70 (p. 2): replace “last accesses” with “last accessed.”
- Line 256 (p. 11): insert a space in “andTAlk” to give “and TAlk.”
- Table 6 caption (p. 17): replace “natural variably” with “natural variability.”
- Line 552 (p. 22): replace “A summary … are presented” with “A summary … is presented.”
- Lines 604, 670, 677, and 685 (pp. 25–27): correct the repeated abbreviation “wRSME” to “wRMSE.”
- Figure 15 caption (p. 31): replace “the NNATL region and the Arctic is excluded” with “the NNATL region and the Arctic are excluded.”
- Figure 19 caption (p. 35): replace “Figure 19.:” with “Figure 19:”.
Citation: https://doi.org/10.5194/essd-2026-496-RC2 -
AC2: 'Reply on RC2', Nico Lange, 15 Oct 2026
Reviewer comments: normal font
Author replies: bold font
General Comment:
This manuscript presents GLODAPv3, a valuable contribution to the oceanographic and broader Earth science communities. The authors have gone beyond data collection by developing and applying methods to evaluate consistency among cruises, identify systematic offsets, preserve temporal trends, and provide additional uncertainty information. These methodological developments strengthen the scientific utility of the product and support the combined use of observations from different periods, regions, and measurement programs. Overall, I consider this an important dataset and a worthwhile contribution to ESSD. My comments primarily concern discrepancies in the reported numbers, consistency between the text and figures or tables, and clarification of several methodological descriptions. Addressing these points would improve the clarity, reproducibility, and usability of the manuscript.
Thank you for your positive assessment of the manuscript and constructive feedback.
Specific Comments:
Lines 600–601 (p. 25) state that salinity data from 152 entries required adjustment, whereas Table 8 (p. 23) reports 158 adjusted entries. The other salinity counts agree, including the 861 entries remaining after exclusions. Please reconcile the adjustment count and ensure that the associated figures use the same final selection. Similar discrepancies occur in the counts reported for other variables; please check all numerical summaries against the final dataset.
The numbers in the text were unfortunately not updated after the final revision of the product. Table 8 contains the correct numbers, whereas the corresponding numbers in Section 3.2 were outdated. We have now updated the text to be consistent with Table 8. The alkalinity paragraph has also been corrected so that the numbers add up.
The seven action categories in Table 4 (p. 10) sum to 1225 for both salinity and oxygen. However, lines 126–127 report 1181 cruise files, and lines 549–551 report 1224 QC entries. Table 4 describes its units as “cruises,” although “entries” may be intended. Please clarify the statistical unit, and use “cruises” and “entries” consistently throughout the manuscript.
Table 4 has been updated and now sums to 1224 entries. The table description is updated accordingly too. Further, the definition of “entries” has been moved to Section 2.1. to be able to consistently apply the two different terms. Note that “cruises” is still used throughout most part of the methods section, while the use of “entry” is restricted to presented results (such as the numbers in Table 4).
Lines 48–50 and 126–127 state that GLODAPv3 contains all GLODAPv2.2023 data plus 57 newly added cruises, yielding 1181 cruises. The cited GLODAPv2.2023 publication reports 1108 cruises; 1108 + 57 = 1165, leaving a difference of 16 cruises. Please reconcile these totals by specifying any additions, reinstatements, removals, or changes in cruise definitions.
Section 2.1 now includes a reference to individual datasets that were previously merged into one cruise in previous GLODAP releases, which explain the mentioned discrepancy. We hope that in combination with the provided v2 cruise numbers in the data product itself, all ambiguities related to differences in cruise numbers are now resolved.
Lines 625–644 (p. 25) distinguish pre-2010, 2010–2018, and post-2018 Japanese silicate data. Lines 895–896 (p. 32), however, refer to preadjustments to pre-2010 and post-2019 Japanese data. These descriptions treat 2019 differently. Please specify the intended year intervals and use them consistently.
We have corrected lines 895-896 accordingly, to now refer to post-2018.
Equation (1) (p. 17) is printed as a sum of Wr/W. This does not express the weighted mean described in lines 385–398 because the denominator lacks the sum of the weights. Please write the numerator as Σj(Wijrij) and the denominator as ΣjWij, and define the summation set. Consider using a symbol such as r̄i for CWMO, rather than x̄, since x already denotes the adjustment vector in lines 344–346.
Equation (1) has been revised to explicitly include the summation over the relevant cruise-pairs offsets, the sum of the cruise-pair offsets weights in the denominator, the cruise index, and the iteration index. We use indexed scalar notation to make the contribution of each crossover explicit. Specifically, (j) denotes the (j)-th entry of the global cruise-pair residual and cruise-pair offset weight vectors, and the summation is performed over the set Ji, i.e. the set of available cruise-pair offsets involving cruise i. Further, we have replaced x by c to avoid confusion.
Lines 349–350 state that shorter time separations receive higher weights. Lines 353–355 state that the additional factors are normalized so that low values indicate high confidence, multiplied together, rescaled to 1–2, and then multiplied by the inverse variance. If the final rescaling preserves the ordering, less confident crossovers would receive larger weights. Please provide the complete scaling equations, including the normalization bounds and whether the final mapping reverses the ordering.
The scaling description has been updated to clarify that the scaling is applied to the standard deviation, which is subsequently used to calculate the (squared) inverse variance. Hence, less confident crossovers receive smaller weights in the FF inversion.
Lines 347–348 define the base weight as the squared inverse of the offset standard deviation. The Figure 5 caption (lines 359–361) instead describes the inverse standard deviation as the weight. Please reconcile the caption and methods with the implemented calculation.
The squared inverse of the scaled offset standard deviation is used as weight. The figure caption is updated accordingly.
Lines 381–382 state that a trend is retained if α exceeds 1.96 times its standard error. Please clarify whether the intended criterion is |α| > 1.96 SE(α), so that significant negative trends are also retained.
Negative trends are also retained as indeed the absolute of α is used as criterion. The text is updated accordingly.
In Figure 10 (p. 23), the legend defines the light-blue, orange, and red categories as <2 × limit, 2–4 × limit, and >4 × limit. The caption (lines 575–577) instead defines them as below the minimum, between one and two times the limit, and above twice the limit. These definitions differ by a factor of two. Please reconcile the legend and caption. Please also specify the year range represented by “2010s”.
The legend is correct, and the figure caption was unfortunately an error. We corrected the caption accordingly and now both legend and caption provide the same information.
The Figure 6 caption (p. 16) states that the figure shows three main processes, whereas the figure labels steps 1–4 and the text also describes four steps. Please make it consistent.
The figure caption is updated accordingly. It now also mentions four steps.
The Table 7 caption (lines 438–440) states that qualifying cruises are adjusted in the third step. Lines 406–415 define step 3 as selection of the “Cruises-to-Adjust” list, while lines 417–419 place the final adjustment inversion in step 4. Please reconcile these descriptions.
The table caption was updated accordingly and now states that the cruises are adjusted in step 4.
Table 10 (p. 33) labels the phosphate row “PO2,” whereas the manuscript otherwise uses PO4. Please correct the variable label.
The label in Table 10 was updated to PO4.
Table 5 and its footnote (p. 12) include pH among the variables receiving additive adjustments. Lines 222–225 and 739–740 explicitly state that secondary QC was not applied to pH in GLODAPv3. Please remove pH from the applicable adjustment list.
Table 5 and its footnote were updated to remove pH. Note that temperature was added.
Table 9 labels salinity as ppm and reports 1.9 → 1.3, whereas lines 604–605 report 0.0019 → 0.0013 for the same wRMSE. Please check the unit.
The provided unit was wrong and Table 9 was updated accordingly.
The reference to Müller and Gruber, “Progression of ocean interior acidification over the industrial era,” is dated 2034 at lines 1287–1288. The paper was published in 2024. Please correct the year.
All references have been checked and a total of 11 references have been updated or added.
Line 227 cites “CO2SYSv2 (Humphreys et al., 2022),” but the corresponding reference at lines 1204–1205 is “PyCO2SYS v1.8: marine carbonate system calculations in Python.” Please verify the software implementation and version intended here and correct the citation accordingly.
Indeed, the incorrect citation was used in the previous draft. We used the MATLAB implementation of CO2SYS, integrated into our Python pipeline using GNU Octave. We have updated the reference accordingly, replacing Humphreys et al. (2022) with van Heuven et al. (2011).
Gruber et al. (2019), cited at line 106, has no matching entry in the reference list. Other citations also appear to be missing. Please cross-check all in-text citations against the reference list and add missing entries or correct the author–year citations as appropriate.
All references have been checked and a total of 11 references have been updated or added.
A challenge is that crossover differences contain both measurement biases and real ocean variability. One possible option would be to sample existing multiyear or multidecadal ocean biogeochemical model output at the actual cruise locations, dates, and depths. A case without imposed measurement biases could test whether the method introduces unnecessary corrections to real ocean changes. Cases with imposed cruise biases and calibration shifts could test its ability to recover known errors. Such experiments could quantify false adjustments, residual errors, regional trend distortion, and performance in weakly connected parts of the crossover network. Where appropriate, they could also test whether corrections inferred from deep water remain suitable for the upper water column. This suggestion does not need to be implemented in the current revision, but it could be discussed in the manuscript or considered for future work.
We thank the reviewer for this suggestion, and we agree that crossover differences can reflect both systematic measurement errors and real ocean variability. A modeling study evaluating how strongly crossover methods might overcorrect would indeed be very beneficial for the community. However, such an evaluation goes beyond the scope of this GLODAP release, where we rely on the established, community-accepted crossover methodology. That being said, we have updated the outlook to include “continuous updating and refinement of the applied methods”.
Typographical Corrections:
Line 70 (p. 2): replace “last accesses” with “last accessed.”
Corrected as suggested.
Line 256 (p. 11): insert a space in “andTAlk” to give “and TAlk.”
Corrected as suggested
Table 6 caption (p. 17): replace “natural variably” with “natural variability.”
Corrected as suggested.
Line 552 (p. 22): replace “A summary … are presented” with “A summary … is presented.”
Corrected as suggested.
Lines 604, 670, 677, and 685 (pp. 25–27): correct the repeated abbreviation “wRSME” to “wRMSE.”
Corrected as suggested.
Figure 15 caption (p. 31): replace “the NNATL region and the Arctic is excluded” with “the NNATL region and the Arctic are excluded.”
Corrected as suggested.
Figure 19 caption (p. 35): replace “Figure 19.:” with “Figure 19:”.
Corrected as suggested.
Citation: https://doi.org/10.5194/essd-2026-496-AC2
Data sets
The Global Ocean Data Analysis Project version 3 (GLODAPv3) – an internally consistent biogeochemical data product for the World Ocean Nico Lange et al. https://doi.org/10.25921/m6tp-mj50
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 524 | 252 | 164 | 940 | 89 | 198 | 158 |
- HTML: 524
- PDF: 252
- XML: 164
- Total: 940
- Supplement: 89
- BibTeX: 198
- EndNote: 158
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
General
GLODAPv3 is the third full re-evaluation of the synthesis, with this time 1181 cruise files, about 1.50 million samples, 1972–2023. Relative to v2.2023 it replaces the WLSQ inversion with the furthest-first (FF) algorithm, lets the regional clusters emerge from the crossover network rather than prescribing them, reduce the minimum adjustment limits (DIC and Talk) by a factor 2, adds per-cruise uncertainties, and adds an explicit trend-preservation step. All these advances are clear. In particular the FF inversion is a striking improvement. Section 4.2 is very insightful, no previous GLODAP release has stated that clearly that the network contains no direct Atlantic-Pacific connection, or that the Indo-Pacific link rests on one cruise. My comments are mostly about the text, not the analysis. The one place where I think the paper could be clearer and more thorough in its explanation is the carbonate system: the choice of input pair, the choice of constants, at a moment when the literature on constants has moved considerably, how users should treat pH data. Nothing below requires new data. Overall, congratulations and thank you on behalf of the community to the authors for putting this together.
Main comments
Minor
Lines 227. Did you mean PyCO2SYS, rather than CO2SYSv2?
Table 5 footnote a still lists pH among adjusted variables.
Table 9 labels salinity in ppm, I am confused if that is a typo or if I am not following something