Drought alters the biogeochemistry of boreal stream networks

Drought is a global phenomenon, with widespread implications for freshwater ecosystems. While droughts receive much attention at lower latitudes, their effects on northern river networks remain unstudied. We combine a reach-scale manipulation experiment, observations during the extreme 2018 drought, and historical monitoring data to examine the impact of drought in northern boreal streams. Increased water residence time during drought promoted reductions in aerobic metabolism and increased concentrations of reduced solutes in both stream and hyporheic water. Likewise, data during the 2018 drought revealed widespread hypoxic conditions and shifts towards anaerobic metabolism, especially in headwaters. Finally, long-term data confirmed that past summer droughts have led to similar metabolic alterations. Our results highlight the potential for drought to promote biogeochemical shifts that trigger poor water quality conditions in boreal streams. Given projected increases in hydrological extremes at northern latitudes, the consequences of drought for the health of running waters warrant attention.

The manuscript reports results of (a) a short term manipulation of a small stream system to induce a hydrological drought and (b) an opportunistic sampling of a natural drought in 2018. The aims of the experiment are valid and worthwhile however I felt the data analysis and interpretation was not of sufficient quality. Specifically, I found the assertions made about anaerobic vs aerobic dominance changes in drought based on CO2 and CH4 concentration ratio changes problematic. Dissolved oxygen is present in the system which does not support assumptions of a switch to system-wide anaerobic dominance and the timescales involved (e.g. average increase of residence time from 38 to 264 min in short term experiment) are very short in relation to typical timeframes of microbial reduction sequences. The potential role of reduced dilution of groundwater fluxes as a mechanism driving change is not explored adequately. Perhaps as a consequence the Principle Component Analysis and some of the other graphs are not convincing about the stated differences between drought and background conditions. Further specific comments are provided below.
Line 34-35 -On hydrological droughts -I feel it would also be worthwhile to mention that hydrological droughts can be caused by human activities independent or semi-independent to climate (e.g. via water diversions or damming of river systems). Line 89 Statement that "Furthermore, measurements from the same locations along this stream reach during the extreme natural drought of 2018 revealed nearly identical relationships between stream chemistry and WRT (Fig. S3)." I disagree with this statement, there is a two order of magnitude difference in ratios at low water residence time.
Line 113 -Statement that "Indeed, analysis CH4:CO2 ratios during the experiment (Fig. S6) indicate that, where reductions in flow were most severe, methanogenesis, the least energetically favorable microbial processes dominated over aerobic metabolism" The surface water shows very little change which does not support this statement. If DO was present as measured, how could methanogenesis occur in surface water? Figure 3 Statement that "drought promoted higher and more dispersed ΔCO2:ΔO2 observation clouds compared to background conditions (Fig. 3), indicating that anaerobic processing of DOM, which generates excess CO2 (Fig. 3a) and CH4 (Fig. 3b) relative to O2, dominated the metabolic balance during drought." I am not completely sure what an "observation cloud" is but I think this refers to the scatter of the data and the authors have drawn a "cloud" around it. In any case there is substantial overlap between drought and non-drought again which leads to some doubts around the stated cause (anaerobic DOM processing) and effects (excess CO2 and CH4) statements made. Also the DO data Figure S5 does not support the statement about anaerobic conditions as mentioned abo ve. The residence time changes involved are relatively short compared with typical times to transition through reduction sequences, particularly to reach methanogenesis.
Hence it may be that groundwater input is less diluted during drought which gives an apparent effect in this regard? I see the authors have done some groundwater measurements, was there CH4 present in groundwater and how could CH4 concentrations in hyporrheic zone and surface water change during drought conditions due to e.g. less groundwater recharge or less dilution in surface water.
Line 125 Statement that "Patterns of hypoxia, as well as its consequences for aquatic life, are well documented for lakes, estuaries and oceans, but are less of a focus in running waters" is incorrect. For example see numerous papers of Baldwin and colleagues for hypoxia in the Murray-Darling river system. The classic research on hypoxia by Streeter and Phelps in the 1930s was developed in running waters.
Figure 4 -why is there such variability in the sensor dataset in 2018, are the authors sure the sensors did not dry out during this time?
A way forward -I found the most convincing results were the stream metabolism data presented in Supplementary Material Figure S4. These show a consistent pattern with residence time and sharp contrast between drought and reference conditions. I would suggest a refocussing of paper to concentrate more on aerobic metabolism changes rather than trying to assert switches to anaerobic mechanisms which are difficult to prove based on the evidence provided (just CO2:CH4 gas concentration ratio shifts and maintenance of oxic-hypoxic conditions).
Reviewer #2 (Remarks to the Author): Review of Drought-induced biogeochemical shifts in high latitude stream networks By Emily S. Bernhardt, Duke University In this MS, Gomez-Gener et al. report substantial shifts in the chemistry, energetics and biogeochemical cycling of surface and hyporheic waters of boreal forested streams as a result of the 2018 northern European drought. Subsequent experimental manipulation of streamflows and water residence time provides strong empirical support for the hypotheses proposed to explain the field patterns observed during the natural drought. This study is among the first to report the consequences of hydrologic drought for high latitude rivers, the results are important and the data and inferences are robust and sound. I would like to see this work published and believe it will be an important contribution to global change biology, ecosystem science and aquatic ecology.
That said, I found the structure of the paper a challenge. While the graphical data presentation is fantastic, the ordering of points and the description of results in the text forces the reader to do too much work and is likely to reduce the impact of the paper. I would like to encourage the authors to consider the following suggestions for increasing the impact of this impressive research.

1) Provide greater clarity about the ecosystem under study and its boundaries.
It is fortunate that I have been to the Krycklan catchments and have some idea of what these systems are like, but this MS does not describe the system under study. It never even mentions what the vegetation is like or anything about the slope or the abundance of wetlands. This context is critically important to the interpretation of your results. Do you believe your findings can apply to all "northern latitude" ecosystems or to all boreal forested wetlands? At present the only description of the study site is that it is Northern Sweden.
It is equally important to be very clear at the outset that this study is about the aquatic ecosystem and not the catchment. Your study ecosystem is a river network and this should be explained. All data come from streamwater and hyporheic samples. Again, the lack of a site description led me to first expect a catchment study and to be initially very surprised by the reduction in aerobic respiration from a wetland dominated catchment upon drainage. You might want to point out this literature -its quite well established that drainage of wetlands and wet soils can enhance aerobic respiration and soil carbon loss. Here you are showing that the aquatic ecosystems draining those catchments may undergo quite different and even opposite trajectories of change.
Providing much greater clarity about the ecosystem to which these questions pertain and to which these findings might be scaled would substantially improve the impact of this paper.
2) Organize the findings in order of their complexity.
I found the presentation of results compelling but disordered. I would suggest that a stronger paper would build from the very obvious to the very exciting.
I would recommend the following flow Drought reduces flow and increases water residence times -please say something about flows here in addition to WRTs -in general it would be useful for the authors to think about fluxes as well as pool sizes. As WRTs increase, dissolved oxygen concentrations decline -many stagnant reaches become hypoxic or even anoxic When we calculate aerobic metabolism from the diel O2 signals we see that rates decline substantially during droughts. Alongside this reduction in O2 and aerobic metabolism, we see increased concentrations of CH4, a shift to high NH4:NO3 ratios and increases in both the CO2:O2 and CH4 to O2 ratios. In the most severe cases, stream reaches shift to methanogenesis as the dominant metabolic pathway. Explain more carefully the work done to document that methanogensis becomes a greater energy source than aerobic respiration… I thought this was THE most exciting finding but the way you determined this from the CH4:CO2 ratios was not explained well enough to allow replication or synthesis 3) More carefully consider whether the higher concentrations of reduced gases indicate an increase in production or a reduction in export… You can get your increase in concentrations via either mechanism and whether total export is changing should matter quite a bit to your interpretations. If the formerthere are landscape/regional implications, if the latter, this simply shifts the local environment but has little impact at larger scales. To rephrase this point, are you observing a catchment scale impact or an aquatic ecosystem only impact? In either case, its worth considering how and whether these in-stream impacts of drought interact with the likely impacts of drought in the uplands. I have a couple of thoughts in this regard Does the composition of DOC inputs to these streams shift as a result of drought? I would expect the declining water tables would lead to more complete degradation of SOM and, perhaps, a reduction in complex DOC molecule export to streams… but perhaps the terrestrial and wetland components of these watersheds simply become disconnected by drought.
What are the net GHG consequences of the decline in aerobic respiration and increase in anaerobic metabolism? Does drought increase or reduce freshwater GHG emissions when you calculate them in CO2 equivalents. I think this is an important question for making this study of broader interest to ecosystem ecologists.
What if anything is known about how the catchment C balance responds to drought? How does this new information contribute to or change that understanding (either in terms of the spatial heterogeneity of responses OR in terms of the total magnitude of ecosystem change).

Minor Comment
Lines 75-77 -I think there is a mistake here… did the WRT increase (as the #s suggest) or decrease (as the wording suggests). Spend a bit more time on this important point (since it is key to several of your graphs).
Reviewer #3 (Remarks to the Author): This paper entitled "Drought-induced biogeochemical shifts in high latitude stream networks" by Gomez-Gener and others reports the effects of a manipulative drought experiment and a naturally dry year on aerobic and anaerobic biogeochemical processes. In both the manipulation and dry year, they found that in drought conditions, reduced compounds, including methane, were more abundant, correlated with water residence time. The study is unique in its spatial extent, demonstrating a network-wide response to drought that will be of broad interest to readers. Additionally, the coupling of a manipulation with a medium-term observational study gives added credibility to the conclusions. I have two minor concerns and several line edits, but after revision, I believe this study would be a valuable contribution to this journal.
1. My main criticism is that the authors attribute all the changes to water residence time without exploring alternative hypotheses for the increase in reduced compounds. How much of the drought effects are due directly to hydrological changes versus other, indirect factors? For example, reaeration of oxygen varies with flow (more turbulence), and temperature is typically warmer during low flow conditions ( 2. This kind of distributed, watershed-scale approach can reveal the extent of change, but can also be leveraged to identify the drivers of change (two recent examples from the Arctic: Connolly et al. 2018;Shogren et al. 2019). Besides looking at the distribution of oxygen concentrations with stream size, I wonder if the authors could explore how different catchment characteristics increase or decrease the likelihood of shifts into anoxia with drought. Said otherwise, could the authors use the rich spatiotemporal data from the Krycklan study area to explore why some rivers experienced persistent anoxia, while others were largely unaffected (e.g. explain the variability observed in Fig. 4)?
Line edits: Line 16: Because neither the experiment nor drought have been previously introduced, it is hard to parse this list. 44-59: This seems too detailed for this portion of the manuscript. Could a more general treatment of the relationship between residence time and biogeochemical conditions suffice here? 62: "Whether" and "if" are yes or no questions. "How much" and "to what degree" are richer and more informative. 75: Typo (increased rather than reduced). Given the subject of the paper, perhaps avoid reduced altogether, except when referring to chemistry (decreased is clearer in this context). 77: How comparable is level of drought severity with amount of flow in the stream? The terrestrial environment, which sets the template for the water chemistry and which becomes more important if upstream flow is cut off as was done here, is not drought stressed, presumably. Lines 82 and onward: The level of detail in the results feels a little imbalanced. Some of them are very specific to this study (PCA axes), while others are more applicable across studies (residence time versus redox state). Reorganizing and perhaps subdividing a little more could improve this. 119 and elsewhere: water chemistry is more specific than water quality 163-164: this seems like a valid hypothesis, but is it based on the results of the current study (as the sentence states) or the cited study? 166-167: Unclear. Potentially rephrase as "CH4 production in aquatic environments could represent a larger proportion of ecosystem-level CH4 balance during these years" or something like that. 168: typo "stimulates" 170: The primary product of denitrification, particularly in situations of long residence time, is the inert gas N2 . This can be an important ecosystem service in nutrient saturated environments, which could be emphasized as a positive tradeoff. 181-191: This seems largely redundant and could be removed without affecting the content of the paper. Figure 4 is difficult to understand. Because it has different axes, the inset graph of the stream order might be more comprehensible as a separate panel.

Reviewer #1 (Remarks to the Author):
The manuscript reports results of (a) a short term manipulation of a small stream system to induce a hydrological drought and (b) an opportunistic sampling of a natural drought in 2018. The aims of the experiment are valid and worthwhile however I felt the data analysis and interpretation was not of sufficient quality. Specifically, I found the assertions made about anaerobic vs aerobic dominance changes in drought based on CO2 and CH4 concentration ratio changes problematic. Dissolved oxygen is present in the system which does not support assumptions of a switch to system-wide anaerobic dominance and the timescales involved (e.g. average increase of residence time from 38 to 264 min in short term experiment) are very short in relation to typical timeframes of microbial reduction sequences. The potential role of reduced dilution of groundwater fluxes as a mechanism driving change is not explored adequately. Perhaps as a consequence the Principle Component Analysis and some of the other graphs are not convincing about the stated differences between drought and background conditions. Further specific comments are provided below.

Response:
We appreciate that reviewer 1 considers this work worthwhile but also value the criticisms of the analysis and interpretation of results. Indeed, these comments prompted us to clarify several issues in the revised manuscript.
Line 34-35 -On hydrological droughts -I feel it would also be worthwhile to mention that hydrological droughts can be caused by human activities independent or semiindependent to climate (e.g. via water diversions or damming of river systems).
Response: We have followed this suggestion and added a sentence describing how droughts can be promoted anthropogenically (lines 32 to 34 of revised manuscript).  Response: We appreciate the interpretation of this analysis provided by the reviewer.
The main goal of using PCA was to get a broad view of how experimental drought influenced the overall chemical patterns in our stream reach (including surface and hyporheic environments). Compared to focus on individual solutes, this approach integrates variation among the electron acceptors and reduced products for which we have data. In the revised version of the PCA, which now includes pH as suggested by R1, the surface water observations between the two treatments do not cluster in almost any case (see Fig. S4 of revised manuscript). Consistent with this, the Wilcoxon Signed-Ranks test shows that surface water PC1 scores were significantly higher during drought. Regarding hyporheic sediments, while we acknowledge some overlap of background and drought samples, it is still apparent that drought promoted the accumulation of reduced relative to oxidized compounds in subsurface environments. For example, the PC1 scores for the hyporheic water was also significantly higher during drought than background conditions. Moreover, PC1 scores for surface and hyporheic water increased non-linearly among sites as drought intensity increased (measured as greater WRT).
Overall, the commentary made by the reviewer, as well as the results obtained from additional analysis stimulated by this comment, made us rethink how we present these results. In the revision, we have: 1) Removed the PCA figure from the main text and placed it in the SI (Fig. S4 of revised manuscript).
2) Only used the PCA results to infer on the distribution of redox-sensitive solutes and gases in the stream and not to directly interpret the relative dominance of anaerobic vs aerobic processes.
3) We have complemented the PCA results with independent analysis of molar NH4 + :NO3and CH4:O2 ratios during the experiment (Fig. 3 of revised manuscript). We have chosen these ratios because they represent most contrasted chemical species in the PCA and provide insight into the potential redox-driven changes in both the nitrogen and carbon cycling.
Finally, we agree with the reviewer that the statement about reduced vs oxidized state might be misleading considering that the redox potential (Eh) was not measured in the experiment. Following his/her suggestion we have i) included pH in the PCA and ii) changed the statements "oxidized status" and "reduced status" by "oxidized solute forms" and "reduced solute forms" to avoid the use of a definition based on the redox state of the system but instead on the redox state of the studied solutes.
Line 89 Statement that "Furthermore, measurements from the same locations along this stream reach during the extreme natural drought of 2018 revealed nearly identical relationships between stream chemistry and WRT (Fig. S3)." I disagree with this statement, there is a two order of magnitude difference in ratios at low water residence time.
Response: This is also a really helpful comment. Although the comparison of CH4:O2 from the same locations between experimental (2017) and natural (2018)

However, this ratio ultimately converged as WRT increased and lateral groundwater inputs declined, suggesting that these chemical signals can be sustained by processes occurring within the stream/hyporheic zone. At the same time, the differences between these two curves across the full range of WRTs does provide an example of how variation in local groundwater hydrology during the onset of drought can potentially exacerbate transitions toward reducing chemical conditions in the stream through hydrologic inputs. To better describe the patterns observed for the CH4:O2 ratios during both the manipulation (2017) and the natural drought (2018), we have now included additional text and two supplementary figures showing:
1) The groundwater inflow dynamics to the stream during the field experiment ( Fig.  S3 of revised manuscript). 2) The relationships between WRT and groundwater CH4:O2 ratios (Fig. 3

of revised manuscript)
Line 113 -Statement that "Indeed, analysis CH4:CO2 ratios during the experiment (Fig.  S6) indicate that, where reductions in flow were most severe, methanogenesis, the least energetically favorable microbial processes dominated over aerobic metabolism" The surface water shows very little change which does not support this statement. If DO was present as measured, how could methanogenesis occur in surface water?  Baker et al., 1999;Jones et al., 2008),

so this is not in itself surprising. In the revised version, we have thus clarified the major role of the hyporheic habitats on controlling the drought-induced biogeochemical changes reported in the manuscript (e.g., see lines 99 to 102; or lines 111 to 113 of revised manuscript).
Figure 3 Statement that "drought promoted higher and more dispersed ΔCO2:ΔO2 observation clouds compared to background conditions (Fig. 3), indicating that anaerobic processing of DOM, which generates excess CO2 (Fig. 3a) and CH4 (Fig.  3b) relative to O2, dominated the metabolic balance during drought." I am not completely sure what an "observation cloud" is but I think this refers to the scatter of the data and the authors have drawn a "cloud" around it. In any case there is substantial overlap between drought and non-drought again which leads to some doubts around the stated cause (anaerobic DOM processing) and effects (excess CO2 and CH4) statements made. Also the DO data Figure S5 does not support the statement about anaerobic conditions as mentioned above. The residence time changes involved are relatively short compared with typical times to transition through reduction sequences, particularly to reach methanogenesis.
Response: To rigorously analyze and interpret results from this stoichiometric analysis (Fig. 4), we have now used three metrics that described the central tendency and the dispersion of ΔCO2:ΔO2. These three metrics are based on a new synthesis of this analytical approach (Vachon et al., 2019) and were computed for surface and hyporheic samples prior to and during the manipulation. We report these metrics, together with a description of how to compute and interpret each, in a new supplementary Table (Table S3 of  Hence it may be that groundwater input is less diluted during drought which gives an apparent effect in this regard? I see the authors have done some groundwater measurements, was there CH4 present in groundwater and how could CH4 concentrations in hyporrheic zone and surface water change during drought conditions due to e.g. less groundwater recharge or less dilution in surface water. Response: See comment above regarding groundwater dynamics during the experiment. Again, to clarify the potential effect of groundwater inputs during the experiment, we have included, described, and discussed both a figure with the groundwater hydrological dynamics during the experiment (Fig. S3 of revised manuscript) as well a figure with the relationships between WRT and stream, hyporheic and groundwater CH4:O2 ratios (Fig.3 of revised manuscript).
Line 125 Statement that "Patterns of hypoxia, as well as its consequences for aquatic life, are well documented for lakes, estuaries and oceans, but are less of a focus in running waters" is incorrect. For example see numerous papers of Baldwin and colleagues for hypoxia in the Murray-Darling river system. The classic research on hypoxia by Streeter and Phelps in the 1930s was developed in running waters. Response: This variability is precisely a consequence of periods with very low flows that promoted episodic declines in stream O2 concentrations during the 2018 summer drought. To better visualize the longitude and magnitude of low O2 periods, daily average of dissolved oxygen saturation has additionally been incorporated to the new version of Fig.5.

Response: We have followed the suggestion made by the reviewer, modified the sentence accordingly and included an additional reference (lines 164 to 166 of revised manuscript).
Also, the high-frequency O2 sensors used in the study (MiniDOT, PME, USA; https://www.pme.com/products/minidot) is an optode that measures dissolved oxygen concentration in water through a fluorescence method. Therefore, exposure of these sensors to the atmosphere (e.g., during episodes of stream drying) should give O2 measurements around the atmospheric equilibrium instead of 0. In addition, we visited these sites frequently during the summer 2018 drought to ensure they remained inundated as the stream network contracted.
A way forward -I found the most convincing results were the stream metabolism data presented in Supplementary Material Figure S4. These show a consistent pattern with residence time and sharp contrast between drought and reference conditions. I would suggest a refocussing of paper to concentrate more on aerobic metabolism changes rather than trying to assert switches to anaerobic mechanisms which are difficult to prove based on the evidence provided (just CO2:CH4 gas concentration ratio shifts and maintenance of oxic-hypoxic conditions).
Response: We have followed the suggestion made by the reviewer and included the aerobic metabolism data in the main manuscript (Fig. 2 of revised manuscript). Additionally, we have re-organized the results to give a greater weight to the droughtinduced effects on the aerobic metabolism patterns. Again, we now interpret the drought-induced patterns as a diversification of metabolic processes or cooccurrence/co-existence of metabolic processes with different redox requirements rather than a change in the dominance. In the revised version, this has been changed consistently throughout the text.

Reviewer #2 (Remarks to the Author):
By Emily S. Bernhardt, Duke University In this MS, Gomez-Gener et al. report substantial shifts in the chemistry, energetics and biogeochemical cycling of surface and hyporheic waters of boreal forested streams as a result of the 2018 northern European drought. Subsequent experimental manipulation of streamflows and water residence time provides strong empirical support for the hypotheses proposed to explain the field patterns observed during the natural drought. This study is among the first to report the consequences of hydrologic drought for high latitude rivers, the results are important and the data and inferences are robust and sound. I would like to see this work published and believe it will be an important contribution to global change biology, ecosystem science and aquatic ecology.
Response: We appreciate the overall positive impression of our study.
That said, I found the structure of the paper a challenge. While the graphical data presentation is fantastic, the ordering of points and the description of results in the text forces the reader to do too much work and is likely to reduce the impact of the paper. I would like to encourage the authors to consider the following suggestions for increasing the impact of this impressive research.
Response: We thank the reviewer for providing such thoughtful and constructive suggestions for how to organize the manuscript to improve its impact. We have restructured the manuscript based on these suggestions (see below for specific responses for the different comments and suggestions raised by reviewer 2).

1) Provide greater clarity about the ecosystem under study and its boundaries:
It is fortunate that I have been to the Krycklan catchments and have some idea of what these systems are like, but this MS does not describe the system under study. It never even mentions what the vegetation is like or anything about the slope or the abundance of wetlands. This context is critically important to the interpretation of your results. Do you believe your findings can apply to all "northern latitude" ecosystems or to all boreal forested wetlands? At present the only description of the study site is that it is Northern Sweden. It is equally important to be very clear at the outset that this study is about the aquatic ecosystem and not the catchment. Your study ecosystem is a river network and this should be explained. All data come from streamwater and hyporheic samples. Again, the lack of a site description led me to first expect a catchment study and to be initially very surprised by the reduction in aerobic respiration from a wetland dominated catchment upon drainage. You might want to point out this literature -its quite well established that drainage of wetlands and wet soils can enhance aerobic respiration and soil carbon loss. Here you are showing that the aquatic ecosystems draining those catchments may undergo quite different and even opposite trajectories of change.
Providing much greater clarity about the ecosystem to which these questions pertain and to which these findings might be scaled would substantially improve the impact of this paper. 2) Organize the findings in order of their complexity: I found the presentation of results compelling but disordered. I would suggest that a stronger paper would build from the very obvious to the very exciting.
I would recommend the following flow: • Drought reduces flow and increases water residence times -please say something about flows here in addition to WRTs -in general it would be useful for the authors to think about fluxes as well as pool sizes.
• As WRTs increase, dissolved oxygen concentrations decline -many stagnant reaches become hypoxic or even anoxic • When we calculate aerobic metabolism from the diel O2 signals we see that rates decline substantially during droughts.
• Alongside this reduction in O2 and aerobic metabolism, we see increased concentrations of CH4, a shift to high NH4:NO3 ratios and increases in both the CO2:O2 and CH4 to O2 ratios.
In the most severe cases, stream reaches shift to methanogenesis as the dominant metabolic pathway. Explain more carefully the work done to document that methanogensis becomes a greater energy source than aerobic respiration. I thought this was THE most exciting finding but the way you determined this from the CH4:CO2 ratios was not explained well enough to allow replication or synthesis.
Response: We thank again the reviewer for suggesting this structure. We have reorganized both the text and figures of the manuscript accordingly.
3) More carefully consider whether the higher concentrations of reduced gases indicate an increase in production or a reduction in export. You can get your increase in concentrations via either mechanism and whether total export is changing should matter quite a bit to your interpretations. If the former -there are landscape/regional implications, if the latter, this simply shifts the local environment but has little impact at larger scales. 2) following the same approach used for Figure 6, we explored relationships between CO2 and CH4 emission fluxes and specific discharges for the five headwater streams in the KCS and during the summer season of the period between January 2010 and October 2018 ( Fig.S8 and Table S4 of revised manuscript).
Our initial prediction was that very low physical reaeration (measured as k600; in m d -1 ) during drought would constrain gas evasion. Thus, CO2 and CH4 emission fluxes would increase with k600 from very low to median flow conditions. However, contrary to these expectations, both CO2 and CH4 emission fluxes remained relatively stable as discharge dropped during very low flow periods. CO2 fluxes responded modestly to increasing runoff (i.e., more inclined slopes) compared to CH4 fluxes. These observations suggest that the transition toward anaerobic over aerobic heterotrophic processes during drought (i.e., higher CH4:CO2 molar concentration ratios) is strong enough to overcome the effect of declining k600 on the CH4 fluxes.
Following the suggestion of the reviewer and the results previously described, we have created a new paragraph to specifically discuss the net GHG consequences of observed biogeochemical responses to drought in the revised version (lines 193 to 213 of revised manuscript).
What if anything is known about how the catchment C balance responds to drought? How does this new information contribute to or change that understanding (either in terms of the spatial heterogeneity of responses OR in terms of the total magnitude of ecosystem change).
Response: This is a good question. Actually, we know that droughts increase the sink strength for CH4 in terrestrial habitats both by limiting its production and increasing its oxidation when water tables are low (Fernner and Freeman, 2011;Strack et al., 2007), a pattern that has recently been confirmed for the 2018 summer drought in the Krycklan (Chi et al., 2019). We have expanded our discussion on the effects of droughts on the overall catchment C balance in northern regions, accounting for our new results in the stream, as well as the results from other publications in terrestrial ecosystems (lines 206 to 213 of revised manuscript).

Minor Comment
Lines 75-77 -I think there is a mistake here… did the WRT increase (as the #s suggest) or decrease (as the wording suggests). Spend a bit more time on this important point (since it is key to several of your graphs).