General comments:

 A complete and robust dataset on the riverine carbon output and variation is definitely interesting to the ocean carbon community and the international stakeholders. According to the title, the data then tend to present would be highly appreciated by many readers and research communities. However, the manuscript itself have some major flaws, which discourage me to recommend to publish it on its current status. The major aspect concerning me including: 1) Language quality is not sufficient and need quite some effort to improve to the level of concise and straightforward academic English. 2) There is a large mismatch between the stated content of the dataset and the actual content of the dataset, specifically, the full carbonate system dataset was said generated, but actually the dataset only consists of the variables DIC and TA. 3) the description the key method , pyCO2SYS v1.7.1 software, for carbon (pCO2, fCO2) estimate is missing.

Specific comments:

1. I did not really understand the processing flow of the data generate, so DIC and TA were estimated with some optimal algorithms and pCO2, fCO2, and pH were the output from the software pyCO2SYS v1.7.1.. but under the title of “full carbonate system data set”, TA and DIC are the primary output and took much of the manuscript and evalution. However, as a reader and a researcher in the ocean carbon community, I would expect pCO2 and fCO2 to be the major variables in the “full carbon system dataset”. Please make this one clear.
2. Throughout the manuscript, the language is not concise or straight forward enough, meaning not academic, and it takes quite some efforts to understand many sentences to grasp their real meaning. And the logic in many of the paragraphs do not really flow.
3. In the abstract, the author mentioned they generated a dataset of full carbonate system, but the variables they mainly present were TA and DIC. So, there should be a statement on the linkage between the variables in the full carbonate system and the TA & DIC. Or there should be a summary on all variable consist of the dataset and their spatial  and temporal resolutions.
4. Line26-28, the uncertainty of the TA and DIC were expressed with absolute RMSE and bias. I suggest the percentage uncertainty should be included, i.e., how much the RMSE and bias account for the minima and minimum of the estimated value of TA and DIC.
5. line 68-69, “Episodic changes in the carbonate system caused by river plumes can result in financial and biodiversity losses and are of paramount interest to local communities, businesses and policy makers (Doney et al., 2020).” , please give specific examples on what kind of financial and biodiversity loess does it cause and how it is of interest to the stakeholders.
6. the first paragraph in section 2.1 is not necessary.
7. line 101-102:” This is a clear weakness of comparing wRMSD values from different sources and across differing regions (Land et al., 2019).” If there is a clear weakness of wRMSD, what is the reason to use it to indicate the quality of the dataset?
8. 7line 175, what is Type A uncertainty?
9. 219- 226, the brief introduction to the software pyCO2SYS v1.7.1 should be included.

10, line 244-250, should be in the method section instead of results.

Technical corrections:

1. line 39-41 “The inorganic carbon content of rivers is poorly constrained due to the difficulties of sampling these highly spatial and temporal variable river outflows.” The logic is not correct, please revise it.
2. line 64-65, “ River plumes can negatively influence wild fisheries and the aquaculture industry (Mathis et al.,2015;Cattano et al., 2018) as plumes can transport low pH waters that can impact the growth and 65 life stages of many marine organisms (Cai et al., 2021) Additionally,” a punctuation  is missing.
3. line 102-103,” Following the methodology of Land, Findlay et al. (2019) we derive RMSDe from wRMSD,”, does not make sense., please revise it.
4. line 120-125, “To be included in the algorithm evaluation, algorithms needed to be applicable within the…… chlorophyll-a.” the sentence is too long to understand, please split it.

The authors reconstructed gridded carbonate system datasets by using a data matchup method based on relationships between carbonate parameters and others which had already been established in the past. While many studies have explored such relationships based on ship-based observations during these decades, the authors utilized these efforts in an effective manner. Such a study is unique and is worth being published, but there are major concerns to be clarified before publication in this journal. I'd like to encourage the authors to improve the study and to revise the manuscript for better understanding.

General comments

Oceanographic characteristics of the studied areas considered, one of the important points of the method is skill to estimate carbonate parameters of low salinity seawaters, which are complexly influenced from both river outflows and heavy precipitation along the ITCZ. On the other hand, relatively higher salinity (S > approx. 34) seawaters in these regions have similar chemical properties to those in the nearest open ocean, where large scale ocean circulations dominate the seawater carbonate chemistry. According to attached supplement files, measurement data used in the matchup process were not necessarily restricted to those of low salinity seawaters. It should be emphasized that the presented method derived more appropriate TA and DIC of low salinity seawaters than others did.

Moreover, secular trends of CO2 were not considered in this study, though time-series reconstructions were addressed. It is needed to show reasonable explanation about that.

Nowadays prevalent machine learning-based methods are used for carbonate system reconstructions; five of the six methods which were cited for evaluating observation-based CO2 sink in the IPCC AR6 assessment used machine learning (Canadell et al, 2021, e.g. Fig. 5.8). It should be explained carefully that this study has some limitation that novel reconstructions cannot be included and legacy of past studies only be used.

Specific comment

Overall

Unnatural uses of brackets “()” have to be checked.

P3 71

Before OceanSODA is presented, successive efforts of investigating empirical relationships between TA/pCO2/DIC and other parameters based on observations have to be mentioned here.

P3 L72-76

A brief explanation of OceanSODA is necessary.

P4 L103

A brief explanation of RMSDe is necessary.

P8 L244- Figure 1

Fig. 1 obviously shows that the four selected algorithms have the lowest RMSDe, but doesn't explain whether they are the best even in low salinity regions. It is questionable that Lee et al. 2000; 2006, which propounded global algorithms and (the latter) didn't use salinity as explanatory variables, have the best skill in low salinity Congo basin. This point should be clarified.

Fig. 4, 5, 8, 9

If DICs were successfully reconstructed, trends of increase in DIC and pCO2 and decrease in pH and Ωs would be also derived. The trends are worth being mentioned in the text to support the validity of this datasets.

Canadell, J. G. et al.: Global Carbon and other Biogeochemical Cycles and Feedbacks. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, pp. 673–816, https://doi.org/10.1017/9781009157896.007. 2021

Lee, K. et al.: Global relationships of total inorganic carbon with temperature and nitrate in surface seawater, Global Biogeochemical Cycles, 14, 979-994, https://doi.org/10.1029/1998GB001087, 2000.

Lee, K. et al.: Global relationships of total alkalinity with salinity and temperature in surface waters of the world's oceans, Geophysical Research Letters, 33, L19605, https://doi.org/10.1029/2006GL027207, 2006.