RC: 1.1 The results show that pCO₂ (and in turn FCO₂) is higher in the larger rivers compared to the smaller rivers, which the authors interpret as resulting from proportional differences in C inputs (both CO₂ and DOC) and metabolism of allochthonous inputs. Since these are connected systems (i.e. the small rivers eventually flow into the larger ones), I'm a bit puzzled how CO₂ would increase downstream due to higher C inputs unless the study design somehow missed high CO₂ inputs from low order streams that directly joined the mainstream?

 AC: Changes in riverine CO₂ can mainly result from the variation of two factors, external C input and in-stream metabolism processes. If the control mechanism in large and small rivers are the same, both are controlled by external C input. Then the higher CO₂ in the large river indicates more C input. Since the large river and the small river are connected, and the CO₂ of the large river comes from the small river, sampling bias could be responsible for the higher CO₂ concentration in large rivers. However, this theory cannot explain the seasonal changes in CO₂ concentration. For small rivers, the highest value of pCO₂ was observed in April, beginning of the flood season, when increased precipitation facilitates the transportation of the soil carbon from land to the river system. However, such an increase was not observed for large rivers in April, although an increase in DOC concentration suggesting more external C input. Instead, a significant increase in CO₂ concentration was observed in July, even though the DOC concentration was slightly lower compared with April. Therefore, we believe that external C input is not the controlling factor, and CO₂ increases downstream due to the high intensity of in-stream metabolism. Long water residence time combined with the high temperature in July facilitated OC decomposition and increased CO₂ concentration in large rivers. However, we do not have direct evidence to support our theory. We are considering using stable isotope to analyze the source of riverine C in the future. When sampling at the river basin scale, it is critical to reducing the error caused by the sampling process. In this study, small and large rivers from eight major sub-basins and different reaches of the mainstream have been sampled. We believe that this sampling strategy can represent the conditions of the basin nicely. There may be small rivers with high CO₂ concentration directly into the mainstream,

but We believe that its impact on the CO₂ concentration of the mainstream is limited due to relatively small discharge.

 

RC: 1.2 Based on Figure 1, it appears that many of the smaller rivers were also at higher elevations. A bias towards higher altitude sites in the smaller rivers could explain the observed trends if these catchments had less vegetation/forest cover and, therefore, fewer C inputs (as both CO₂ and DOC). Indeed, the authors observed higher DOC concentrations in larger rivers, which they assume fuels higher respiration. Where does this DOC come from if it doesn't pass through smaller rivers first? I suspect there is some sampling bias at hand.

 AC: We agree that small rivers sampled tend to have a slightly higher elevation than large rivers. It is also partially because those headwater streams tend to be distributed in higher elevation in the hill-dominated Dongjiang river basin. Land cover will indeed influence the C input, and we will address its impacts in the discussion. However, small catchments in higher elevations tend to have more forest cover, so less C input may not be responsible for the low CO₂ in small rivers. Even though a slightly higher average DOC concentration was observed in large rivers than small rivers, the difference is not statistically significant. The DOC concentrations of the two are similar. That is why we believe higher CO₂ concentration in large rivers is due to favorable decomposition conditions rather than more supply of OC. It is difficult for OC to convert into CO₂ in small rivers due to the high flow velocity and short water residence time; thus, it could be transported and fuel the heterotrophic respiration in large rivers. There may be small rivers with relatively high C concentrations that could directly join the mainstream, but the impact on C input should be minor due to relatively small discharge.

 

RC: 1.3 There are additionally more processes, such as photo-oxidation or titration of the carbonate equilibrium via organic acids (indeed, you see increasing CO₂ with decreasing alkalinity), that could impact some the observed downstream increase in CO2. These aspects are not discussed in the manuscript and the authors conclude too strongly that they know the responsible drivers without data to support such claims. Since more highly productive vegetation in the catchment could result in both higher CO2 inputs and higher DOC that fuels respiration, I think it would be useful to explore the relationship between C concentrations (pCO2 and DOC) and catchment land-cover (perhaps as a fraction of wet area, similar to Rocher-Ros et al 2019, L&O Letters or % forest cover).

AC: Indeed, we have concluded too firmly about the responsible drivers without enough direct evidence. Other factors may affect the concentration of CO₂ and should be discussed. We will reduce certainty and explore the relationship between C concentration and land cover.  We will also discuss the potential impact of more processes. For example, photo-oxidation may be responsible for some of the deviation of ΔCO2:ΔO2 stoichiometry line.

 

 RC: 2.The discussion of spatial and temporal patterns is blended together and needs to be disambiguated a bit. It is hard for the reader to make sense of these various overlapping trends. I would suggest starting with one (spatial), then the other (temporal) before finishing on how they overlap to result in the observed pattern.

 AC: Thank you for your advice. We will revise the discussion about spatial and temporal patterns and improve the referencing of figures and tables.

 

RC: 3.Increased precipitation can both increase the transport of terrestrial C (including CO2) and dilute it. How do you know which process dominates?

 AC: We estimate the intensity of those two effects by analyzing the temporal pattern of riverine CO₂. For example, precipitation and CO₂ concentration increased simultaneously from January to April in small rivers, suggesting that the increase of terrestrial C transportation is the dominant process. In comparison, precipitation was similar between April and July, but CO₂ concentration decreased during this period.  The dilution and depletion effect caused by precipitation should be more important in this case.

 

RC: 4.Throughout the discussion, the authors fail to reference their figures or tables in many cases that would make it much easier to observe their explanations.

AC: Thank you for your advice. We will revise it in the manuscript.

 

RC: 5.Given the high resolution of the pCO₂ data, would it not be interesting to upscale outgassing for the whole basin? Perhaps it could be compared to DOC/POC export if those have been previously estimated (or even roughly estimated using your values). A the very least, I think the authors' data could be nicely displayed on a map (Similar to Figure 1 of Rocher-Ros 2019, Limnology and Oceanography Letters)

 AC: We are also very interested in the calculation of basin-wide CO₂ emission. After all, one of our objectives is to provide support for more accurate global CO₂ emissions estimates. The estimation of CO₂ emissions at the watershed scale is not only limited by the accuracy of the CO₂ data but also the accuracy of the river network extraction. Currently, we are working on a  study about the basin-wide CO₂ emissions estimates in the Dongjiang river basin. We intend to perform higher-precision river network extraction and water area calculation by combining remote sensing images and DEM. A more accurate watershed-scale CO₂ emissions estimation will be carried out, and its relationship with lateral carbon transport and net ecosystem productivity. This study intends to focus on the difference in carbon emissions between large and small rivers.

 

RC: Overall, I think the discussion of the drivers of CO₂ variability is overstated. Specifically, there is no direct evidence of lateral soil CO₂ nor dilution effect caused by precipitation. There doesn't seem to be much of a difference in dCO₂ vs. dO₂ between large and small rivers (Figure 6), suggesting that metabolism is similar. At minimum, the current discussion would need to justify why simultaneously low DOC and CO₂ are not an artifact of altitude/land-cover.

AC: Indeed, we have concluded too strongly about the responsible drivers without enough direct evidence. Other factors may affect the concentration of CO₂ and should be discussed. We will reduce certainty and discuss the potentials impact of more processes. Regarding dCO₂ vs. dO₂, dO₂ is higher in small rivers than large rivers with the same dCO₂, suggesting that the impacts of factors other than metabolism should be more obvious in small rivers. As for the relationship between DOC and CO₂, we will further elaborate on why  DOC may not be the controlling driver of the CO₂ changes and discuss the possible influence of altitude and land cover.

 

RC: 16-17 - what direct evidence of soil CO₂ and dilution is there to support this statement?

AC: Due to the lack of stable isotope analysis, we do not have direct evidence pointing to the source of carbon. We can only estimate the effects of soil CO₂ input and dilution by analyzing the temporal pattern of riverine CO₂. For example, precipitation and CO₂ concentration increased simultaneously from January to April in small rivers, suggesting an increase in terrestrial C transportation. Meanwhile, a more rapid response of riverine CO₂ to terrestrial C input in small rivers comparing with large rivers in April suggests different C sources controlling the CO₂ changes. We will reduce certainty and discuss the limitation of our estimation.

 

RC: 96 - Figure 1 could be supplemented with a land-cover map. Many of the smaller rivers appear to be at higher elevations and I am curious if they are less forested.

AC: We will add a land-cover map along with the spatial pattern of CO₂ concentration. Actually, catchment at higher elevation in the Dongjiang river basin is more forested, we will evaluate the impact of land cover on riverine CO₂ concentration in the discussion.

 

RC: 103 - Figure 2's data might be better suited for a bar graph?

AC: Thank you for your advice. We will revise it in the manuscript.

 

RC: 163 - I think the reference to equation 2 is incorrect here.

AC: Thank you for your advice. We will revise it in the manuscript.

 

RC: 195 - There is no hydrologic data in Table 1. Discharge should be presented.

AC: Thank you for your advice. We will provide related hydrologic data in the supplement.

 

RC: 197 - Again there is no stream width or discharge data presented anywhere in the manuscript besides these lines of text. 

AC: Thank you for your advice. We will present stream width in the supplement.

 

RC: 202 - U10 is undefined.

AC: Thank you for your advice. U10 has been defined in L125.

 

RC: 275 - DOC and CO₂ can simultaneously be transported from terrestrial systems, which also might explain their correlation.

AC: We agree that DOC and soil CO₂ can simultaneously be transported from terrestrial systems. However, a discrepancy in the temporal pattern of DOC and riverine CO₂  was observed, which suggests that DOC input might not be the controlling factor. That's also why we discuss the impacts of internal metabolism according to the result of dCO₂ vs. dO₂.

 

RC: 297-318 - This section is very overstated and not the only way to interpret these data. I recommend revising and rephrasing to reduce certainty and include alternative explanations.

AC: Thank you for your advice. We will rephrase to reduce certainty and discuss the possible impacts of land cover.

 

RC: 381-382 - This is possible, but not certain.

AC: Thank you for your comment. We will reduce certainty here.

 

RC: 390 - Respiration and photosynthesis can occur simultaneously.

AC: We agree that respiration and photosynthesis can occur simultaneously,  And we are interested in the intensity of those two processes in the Dongjiang River. In the nearby Xijiang River, high DO and CO₂ occurred simultaneously in summer, indicating that photosynthesis is dominant and C source other than respiration should be responsible for high CO₂ concentration observed. In contrast, DO and riverine CO₂ were negatively correlated, and supersaturated CO₂ was observed in the Dongjiang River, indicating that the effect of respiration is more obvious.

 

RC: 405 - The units for pCO₂ are not consistent (sometimes uatm sometimes ppm). What about Borges 2015, nature geoscience that includes a significant amount of data for rivers in central Africa? Also Mann et al. 2014 JGR-Biogeosciences has additional pCO₂ data. Lastly, is the Mississippi River really a subtropical basin?

AC: In some studies, the results of pCO₂ were only provided in ppm. We will add notes under the table. Thank you for the recommendation. We will add extra data from Africa. According to the Köppen Climate Classification system, sampled lower Mississippi river basin belongs to the Humid subtropical climate zone.

 

RC: 409-412 - Again, I don't think these conclusions are justified

AC: Thank you for your comment. We will reduce certainty here.

 

RC: 417- Still don't really see how depletion would only affect small streams and not the larger ones they flow into?

AC: If lateral C input is the primary driver of riverine CO₂ in small and large rivers, depletion should affect both small and large rivers. However, the decrease in riverine CO₂ during the wet season was only observed in the small rivers, not the large rivers, indicating different controlling factors. One possible explanation is that, for large rivers, DOC concentrations in April and July are both more than enough to support the requirement of respiration. Therefore, the intensity of in-stream metabolism rather than DOC concentration controls the temporal pattern of riverine CO₂ in large rivers. We understand that this is only one possible explanation without enough direct evidence. Therefore, the limitation of this explanation will be discussed in the manuscript.

 

RC: My one main concern is in relation to the way the dataset was collected. It appears that the data was not collected with replicates and in a kind of snapshot approach across a large river basin. It is therefore very challenging to standardise for hydrological conditions, time of day etc. when co-ordinating sampling from such a large basin. However, there should be some analyses and discussion around this point to explore how this might impact the results as presented. One of the aims of the study is to “investigate the spatial and temporal pattern of pCO2 and CO2 emission along stream size spectrum” – how would different sampling conditions affect this? Not enough context is provided to reassure the reader that artefacts of the sampling process are not driving at least some of the variability observed in this dataset.

AC: Intending to improve the representativeness of our dataset, we have carefully chosen the location and time of our fieldwork campaigns. In this study, six reaches from the mainstream of the Dongjiang River and 37 rivers from eight main sub-basins were sampled. In total, five fieldwork campaigns were performed during different hydrological conditions, including two campaigns in the dry season and three in the wet season, and each campaign took about 2 weeks.  Even so, we agree that sampling conditions could affect the result, and we will discuss the possible impact of the sampling process in our revised manuscript.  As for data collection, the measurements of pCO₂ and FCO₂ at each site were repeated twice, and the average was then calculated and used in this study. Meanwhile, the variation between the two measurements was less than 5%, and extra measurements will be conducted if the difference is larger than 5%. Detail about the measurement will be further elaborated in the manuscript.

RC: L55. Please indicate which references refer to which so that the reader can use this as a pointer towards specific studies which observed one or the other pattern.

AC: Thank you for your advice. References will be changed accordingly.

RC: L96. What type of forest? Just to clarify, these “plains and hills” are predominantly covered by this forest? Please provide some more information on the extent of coverage.

AC: Overall, about 67% of the catchment is covered by evergreen Broad-leaved Forest. More detail will be provided in the manuscript. 

RC: L127. Please provide details of the flow meter, including accuracy etc.

AC: Flow velocity was determined by using a Global Water Flow Probe FP111 with a precision of 0.1 m s^-1. Detail will be provided in the manuscript.

RC: L130. Can you provide an indication of how big this underestimation might be? An order of magnitude, or just a few percent?

AC: Flow velocity measured near the bank is about 60% to 80% of that in the middle of the river. Detail will be provided in the manuscript.  

RC: L142-3. These volumes are larger than what are typically used for headspace extractions. Did you test this method for accuracy compared to smaller volume methods or can you provide a reference to back up this approach? Mostly to confirm that full equilibration between water and headspace is occurring within 1 min of shaking.

AC: Large headspace has been used in previous studies. For example, a 600ml conical flask has been used by Müller et al. (2015), and we also used large volume headspace in our previous study (Ran et al., 2017). According to our test, shaking the flask for one to five minutes could yield a similar result if we shake it vigorously (200 times per minute). Related references will be mentioned in the revised manuscript.

RC: L162. Think this is supposed to be eq 3.

AC: Thank you for your advice. We will revise it in the manuscript.

RC: L189. Does this include replicates at any sites? Or were single measurements only of FCO2 and pCO2 undertaken at each site? It seems strange to omit any kind of replication at each measurement site, so I would encourage the authors to explain why and discuss whether this lack of replication had any major impact on their findings. Further, were sites measured all in the same day or over multiple days? If so, how might time of day or hydrologic conditions varied across these measurements within each campaign? I know you can’t go back and fix any of these potential issues after the fact, but some discussion of potential issues here would be useful to convince the reader that there these decisions made when designing the sampling strategy have not substantially impacted the data that is presented here. This is most concerning when I look at Table 1. The values appear very consistent across all the sites, yet the standard deviation compared to the means are very large in some cases.

AC: The measurements of pCO₂ and FCO₂ at each site were repeated twice, and the average was then calculated. Meanwhile, the variation between the two measurements was less than 5%, and extra measurements will be conducted if the difference is larger than 5%. In total, five fieldwork campaigns were performed during different hydrological conditions, including two campaigns in the dry season and three in the wet season, and each campaign took about two weeks. During the dry season, the hydrological condition was relatively stable due to the lack of precipitation. In contrast, the hydrological conditions could vary during the wet season. However, we believe that it may not drive the spatial pattern observed since both small and large rivers have been sampled before, during, and after precipitation events. As for the time of day, all measurements were conducted during the daytime, so the CO₂ concentration might be slightly lower than the night time. However, it may not have major impacts on the spatial and temporal pattern observed. Because Chl a in the Dongjiang River is relatively low, so as the impacts of photosynthesis on riverine CO₂. Nevertheless, We will discuss the possible impact of the sampling process in our revised manuscript. 

RC: L197. Change Q to “discharge.”

AC: Thank you for your advice. We will revise it in the manuscript.

RC: L225. What did this “strongest increase” actually relate to? Stream order is just a proxy for many things, including discharge, catchment characteristics etc. This is not fully discussed or addressed in the discussion.

AC: Initially, we are looking forward to a gradual decrease or increase in pCO₂ from low order to high order streams. However, we have observed a strong increase in pCO₂ from the third order stream to the fourth order stream. Meanwhile, first to third order streams have similar pCO2 values, and fourth to seventh order streams have similar pCO2 values. That's why the discussion mainly focus on the difference between small rivers (first to third order streams) and large rivers (fourth to seventh order streams) in pCO₂, k600 and FCO₂. All discussion about small and large rivers are actually related to stream order.

RC: L245. Clarify the sentence here: “indicating that the majority of the river network is a carbon source”.

AC: Thank you for your advice. We will revise it in the manuscript.

RC: L259. Not sure what this means, between wet vs dry seasons?

AC: Among five fieldwork campaigns, two of them were performed in the dry season, and three of them were performed in the wet season. Here we have compared the result between different campaigns in the same season. We will clarify it in the manuscript. 

RC: L264. High compared to what?

AC: It is high compared with other months. We will clarify it in the manuscript.

RC: L273. This too broad a statement to really be useful. This is dependent on the river and its setting etc. Perhaps rethink the purpose of this opening sentence and target it more directly to the immediate discussion.

AC: Thank you for your advice. We will revise it in the manuscript.

RC: L310. Which “should” lead to a decrease? Because you then observed pCO₂ to increase, rather than be diluted.

AC: Decrease in pCO₂ was observed in small rivers from April to August. It could result from dilution and depletion effects. 

RC: L322. Decomposition of organic carbon “within the water column” (internal DOC decomposition)?

AC: Thank you for your advice. We will revise it in the manuscript.

RC: L331. Plenty of studies have indicated that DOC can be readily decomposed in headwater streams, e.g. Vonk et al. 2013 (doi: 10.1002/grl.50348), Dean et al. 2019 (doi:10.1029/2018JG004650).

AC: Thank you for your recommendation. Indeed, DOC can be readily decomposed in some headwater streams, but it also depends on their setting. Headwater streams in the peatland or permafrost region tend to have a low gradient and, thus, more favorable conditions for DOC decomposition. In contrast, headwater streams in the Dongjiang river basin usually have a high gradient and high flow velocity due to a predominantly hilly landscape. Therefore, it will be more difficult for DOC to be decomposed here. Nevertheless, it might be a good idea to include the result of those two studies in our discussion to support our arguments.

RC: L334. Should you not then see a correlation between DOC and pCO₂?

AC: Yes, higher DOC and pCO₂ were both observed in the wet season comparing to the dry season. However, we also found that changes in DOC and pCO₂ were not simultaneously. For example increase in pCO₂ was not observed for large rivers in April, although an increase in DOC concentration suggesting more external C input. Instead, a significant increase in CO₂ concentration was observed in July, even though the DOC concentration was slightly lower compared with April. Therefore, favorable conditions for OC decomposition might be more critical for the increase of pCO₂ compared to the concentration of DOC.

RC: L343. In line with previous studies, e.g. Long et al. 2015 (doi: 10.1002/2015JG002955). Fig 8. I suggest repositioning the legend so that single blue dot is more obvious.

AC: Thank you for your recommendation. We will include the reference in our discussion and reposition the legend. 

RC: L393. For all rivers? Or large rivers? Because the earlier discussion suggested internal production of CO₂ was more important for the larger rivers.

AC: We believe that the internal production of CO₂ was more important for the large rivers. However, for small rivers, respiration could also contribute to high CO₂ in the summer, even if it’s not the primary driver. Meanwhile, high DO and CO₂ occurred simultaneously in summer in the nearby Xijiang river, indicating that photosynthesis is strong and C source other than respiration should be responsible for high CO₂ concentration. We will rephrase the sentence to clarify our arguments.

Reference

 Müller, D., Warneke, T., Rixen, T., Müller, M., Jamahari, S., Denis, N., Mujahid, A., and Notholt, J.: Lateral carbon fluxes and CO 2 outgassing from a tropical peat-draining river, Biogeosciences, 12, 5967, 2015.

Ran, L., Lu, X. X., and Liu, S.: Dynamics of riverine CO2 in the Yangtze River fluvial network and their implications for carbon evasion, Biogeosciences, 14, 2183-2198, https://doi.org/10.5194/bg-14-2183-2017, 2017.