Autumn destabilization of deep porewater CO2 store in a northern peatland driven by turbulent diffusion

The deep porewater of northern peatlands stores large amounts of carbon dioxide (CO2). This store is viewed as a stable feature in the peatland CO2 cycle. Here, we report large and rapid fluctuations in deep porewater CO2 concentration recurring every autumn over four consecutive years in a boreal peatland. Estimates of the vertical diffusion of heat indicate that CO2 diffusion occurs at the turbulent rather than molecular rate. The weakening of porewater thermal stratification in autumn likely increases turbulent diffusion, thus fostering a rapid diffusion of deeper porewater CO2 towards the surface where net losses occur. This phenomenon periodically decreases the peat porewater CO2 store by between 29 and 90 g C m−2 throughout autumn, which is comparable to the peatland’s annual C-sink. Our results establish the need to consider the role of turbulent diffusion in regularly destabilizing the CO2 store in peat porewater.

13. Line 327-329: I think this calculation needs to take into account the approximately 10 times higher solubility of CO2 relative to CH4 in these conditions so CO2 is likely much lower than predicted here. 14. Line 439: should this be >2meters? 15. Line 440-442: I don't understand this. The deepest porewater (2m) was sampled once a month, but 2.5m was sampled hourly in some months? This part needs clarification. 16. Paragraphs 522-552: I've done some diffusion modeling and I can't quite follow what they are doing here. For example, Is the mass transfer method (e.g. Happell et a., 1995) or the boundary layer method used to calculate exchange coefficient here? If the latter, the result is very sensitive to boundary layer thickness. How was that chosen? I think this section could use some clarification. (see: MacIntyre, S., Wanninkof, R. and Chanton, J.P., 1995. Trace gas exchange across the air-water interface in freshwater and coastal marine environments, in, edited by: Matson, PA and Harriss, RC, Biogenic trace gases: Measuring emissions from soil and water. Methods in ecology, 52-97. Happell, J.D., Chanton, J.P. and Showers, W.J., 1995. Methane transfer across the water-air interface in stagnant wooded swamps of Florida: Evaluation of mass-transfer coefficients and isotropic fractionation. Limnology and Oceanography, 40(2), pp.290-298.
Reviewer #2 (Remarks to the Author): The manuscript presents evidence of a possible source of carbon dioxide losses from peatlands that has been ignored previously. The authors argue that turbulent diffusion has the potential to move deep porewater CO2 closer to the peatland surface, where it could be released by atmospheric emission and hydrological export. Their estimates of autumn loss of porewater CO2 storage for a Swedish peatland are comparable to the peatland's annual C sink. The authors challenge the perception that greenhouse gases in deep peat porewater are largely inert and raise the possibility of substantial errors in global estimates of peatland C sinks and GHG emission. The results are original and should be of interest to peatland researchers as well as to researchers with wider interests in the terrestrial carbon cycle and greenhouse gas emissions.
The key argument is supported by analysis of four years of near-continuous measurements of porewater CO2 and temperature at varying depths in a peatland in northern Sweden. The authors used a heat budget approach to estimate the turbulent diffusion coefficient for CO2. The temporal extent and resolution of porewater CO2 data are rare, if not unique to this study. The sampling strategy is valid for demonstrating recurring thermal instabilities and associated seasonal fluctuations in porewater CO2 concentrations and storage. Previous studies based on mineral solute profiles (e.g., Griffiths and Sebestyen 2016) have already demonstrated that deep porewater is dynamic rather than static. The unique contribution of this study is to demonstrate this phenomenon for dissolved CO2 and to propose that these seasonal porewater dynamics are explained by turbulent diffusion.
The section of the manuscript that considers the implications of turbulent diffusion for the peatland C budget is less convincing. The authors argue that rapid rise of deep porewater to the surface could release CO2 to the atmosphere and/or to runoff. They estimated the cumulative porewater CO2 loss through autumn as comparable to the peatland's annual net ecosystem exchange and an order of magnitude larger than its stream CO2 export, and the daily flux of deep porewater CO2 toward the surface during periods of instability as comparable to mean annual rate of ecosystem respiration. The difficulty is that the measurements of atmospheric net CO2 exchange and stream CO2 export do not show the pattern or magnitude of fluxes that would be expected. To account for this lack of evidence for rapid efflux of CO2, the authors argue that the method of partitioning NEE into GPP and Reco is flawed. Further, they argue that their peat depth profile may be unrepresentative of the whole catchment, explaining why autumn pulses of stream CO2 export were negligible compared to losses of porewater CO2 storage.
The possibility of substantial losses of deep porewater CO2 is tantalising, but the evidence for its rapid efflux to the atmosphere or stream water is weak: based on the data available, the porewater CO2 budget is incomplete. It would be helpful to outline what measurements and experimental set-ups would be required to confirm and quantify C losses from deep porewater. For example, peat profile monitoring of hydraulic heads and the concentrations of a conservative solute would help to disentangle the contributions to vertical mass transport of advection, molecular diffusion and turbulent diffusion. Tracer studies would help to establish the fate of 'lost' porewater CO2 and, ideally, CH4.
Reference Griffiths, N.A., and Sebestyen, S.D. (2016) Dynamic Vertical Profiles of Peat Porewater Chemistry in a Northern Peatland. Wetlands 36: 1119-1130. DOI 10.1007 Specific comments: Line 36 -reference needed for 'hydro-physically inactive' Line 41 -Does 'CO2' refer to concentration or stock? Spelling: 'instantaneous' Line 60 -'manual characterization with a low time resolution' -rephrase for clarity, e.g., periodic manual measurements? Line 111 (and following) -'overlying' Lines 120-121 -Is the second value for the 1-2 m layer (not 1.5 m)? Lines 124-126 -Near-surface porewater CO2 concentrations will also be affected by rhizosphere processes, which change seasonally. Line 195 / Figure 5b -Given the relationship stated in the equation, it would make more sense for Kz and the trend lines to be plotted against 1/N2 rather than N2. Overview This paper is a thought-provoking account of a high-quality dataset from a northern peatland that suggests that CO2 dissolved in deep (> 0.5-0.75 m depth) porewater can be lost from peatlands during the autumn. Previously, this store of CO2 has been assumed to be largely static or having a slow turnover. In a mostly well-written paper, the authors suggest that their dissolved CO2 data can be explained by the process of turbulent diffusion, which can lead to rates of dissolved CO2 transport that are orders of magnitude greater than molecular transport. In a careful analysis, the authors also consider the importance of the posited process compared to other processes in the carbon cycle in peatlands. Overall, I enjoyed reading the paper and I believe there is information in it that is very worthy of publication in Nature Communications. However, I struggled to understand parts of the manuscript and I believe the authors could do a better job of explaining their proposed mechanism. Below, I explain how I think the paper could be improved. I have also marked up the pdfs of the manuscript and the SI, and the editor and authors are referred to these for other comments and concerns that I recommend are addressed.
More than one type of turbulent diffusion? Groundwater scientists usually consider there to be two ways by which solute mixing may occur: molecular diffusion and mechanical dispersion, the latter being analogous to diffusion but occurring in laminar flows due to the 'splitting' of flow through a porous medium as flowing water 'samples' the wide range of pore sizes. Turbulence can sometimes occur in porous media in, for example, biogenic macropores such as wormholes, larger voids between soil peds, and joints in bedrock. Turbulent transfer of a range of properties is also known to occur in surface waters, and the transfer has been found to follow mathematically the diffusion equation; hence, the use of the term 'turbulent diffusion'. The latter occurs when there is rapid mass transfer of water. The authors, however, seem to be suggesting a type of turbulent transfer that does not involve mass flow of water. From the text and Figure 6 they propose that pore-scale turbulent eddies form that help transfer dissolved CO2 along a concentration gradient. They also note that this turbulence causes a transfer of CO2 that is orders of magnitude greater than would occur via molecular random walks.
In principle, I find this suggestion plausible, but what I find confusing is the authors' explanation, or lack of explanation, of the mechanisms involved. In short: what causes the turbulence? The authors are clear in suggesting mass flow is not responsible (e.g., on lines 226-7), and they warn against treating the porewater domain in the same way as lake water where turbulent diffusion has also been observed. Perhaps it is my ignorance showing here, and perhaps other scientists will clearly understand what the authors mean. However, I suspect many target readers of the paper will be like me and not know how turbulent diffusion can occur without mass flow. I guess part of my problem here is that I struggle to see how turbulent mixing at the scale posited by the authors can occur in the pores found in the deeper peat at their site. They suggest that turbulent eddies in the peat pores might form at scales of mm to cm but I don't see how that is possible because most pores in the deep peat will be of mm scale or much less (based on the hydraulic conductivity data presented in the SI). Notably, on lines 227-235, the authors themselves seem somewhat uncertain about how turbulent diffusion can occur in peat in the absence of mass flow.
Ebullition and measurement of dissolved CO2 I wonder too whether bubble formation and transport within the peat might also offer a part explanation of the authors' findings. Biogenic gas bubbles tend to accumulate in peat during the growing season and reach a peak unit volume by late summer and early autumn. These bubbles contain CO2 as well as CH4 and their movement upwards through the peat profile may also facilitate CO2 transfer. Their movement through, and loss from, the peat profile will also cause mixing of pore water that might give rise to turbulent diffusive transfer of dissolved CO2. I should add that sudden episodes of gas bubble loss from peat are not always easily detected or measured by eddy correlation. The flux tower used by the authors was also somewhat distant from the monitored peat profile so may has missed ebullition events near the sensors.
I have a slight concern about the measurement of dissolved CO2. What happens to the CO2 sensors if bubbles are in contact with the sensor membrane? Is it possible that the sensors measure a combination of dissolved and free-phase CO2? Also, if bungs were fitted to the tops of the sampling tubes, how did flow of water into and out of the tubes occur? Such tubes need to be vented to allow water ingress and egress.

Response to Reviewers
Reviewer #1 (Remarks to the Author): Summary: This paper describes thermal destabilization in a peatland which the authors suggest contributes to increasing flux of CO2 from deep peat porewater in the autumn. The high rate of flux is comparable to the C sink of this peatland suggesting a mechanism for a potentially strong peatland-climate feedback.
General Impressions: I generally like the idea of thermal destabilization as a mechanism for gas transport in peatlands. I think the authors could consult more of the recent peatland literature to better contextualize their study and results. The methods lack clarity in a few points specifically with regards to the flux calculations (detailed comments below) that I think needs to be revised before the study results can be evaluated fully.
1. I am interested in the idea of rapid diffusion of CO2 facilitated by thermal instability, but I'm not convinced about the opening premise that deep CO2 in peatlands is generally considered to be inert. Several of the modelling groups explicitly model diffusion and ebullition of CO2 from deep towards the surface (e.g., Ma et a., 2017). Ma, S., Jiang, J., Huang, Y., Shi, Z., Wilson, R.M., Ricciuto, D., Sebestyen, S.D., Hanson, P.J. and Luo, Y., 2017. Data constrained projections of methane fluxes in a northern Minnesota peatland in response to elevated CO2 andwarming. Journal of Geophysical Research: Biogeosciences, 122(11), pp.2841-2861 We agree with the reviewer that this was an overstatement. Many studies have documented considerable dynamics in groundwater flow, with significant implications for peat solute and gas concentrations. The peat porewater CH4 store, in particular, is known to be dynamic due to the role of ebullition. We nonetheless note a general agreement in the literature that vertical transport of porewater CO2 (a more soluble gas than CH4) is governed by molecular diffusion, which is a slow and constant removal process (Line 46-49).
We have revised the introduction to instead lay the concept of acrotelm and catotelm as two main layers of a peatlands, and detail their roles in peatland CO2 cycle (atmospheric and hydrological fluxes) (Line 28-44). We avoid stating that the catotelm is inert, and rather detail its role in the peatland CO2 cycling (which is lower than that of the acrotelm). We subsequently state the vertical transport of catotelm porewater CO2 towards the acrotelm is considered to take place by molecular diffusion, a slow and constant process , where we also refer to the modelling studies that you recommended (Ma et al, 2017) and (Walter et al, 2000). The previous version of the introduction listed multiple transport pathways (mass flow, diffusion, ebullition) for both CO2 and CH4 in peat porewater, which we consider was out of focus. Our study specifically challenges the assumption of a slow and constant rate of vertical diffusion in the catotelm, not the effect of other processes, which have been studied in more detail. These other hydrological processes, and their possible role in the peat porewater CO2 dynamics, are instead detailed and discussed in the results & discussion section (e.g., Line 186-204, and section "Other drivers of porewater CO2 dynamics (Line 221-270").
2. Line 30-31: The authors state that the ability of peatlands to act as a sink is attributable to processes in the oxic surface. I would argue that is not the case and the sink ability of peatlands occurs when peat is buried in the catotelm away from energetic TEAs and protected from high rates of microbial decomposition by numerous processes including cold, low pH, and accumulation of potentially inhibitory compounds. The ability of peatlands to act as a sink depends on the balance between primary production forming peat and the ability of microorganisms to remineralize. If PP is high relative to decomposition, the peatland is a sink, if PP is low relative to decomposition, then they act as a source.
2 Z. F., D. I. Siegel, S. S. Dasgupta, P. H. Glaser, and J. M. Welker (2014), Stable isotopes of water show deep seasonal recharge in northern bogs and fens, Hydrological Processes, 28(18), 4938-4952, doi:10.1002/hyp.9983. We now present a more detailed evaluation of the potential role of changes in mass flow to explain the losses of porewater CO2 in autumn , where we also cite the references, you recommended. 4. Line 37-39 This statement needs some references. This statement has been removed because we judged it was not relevant to introduce the study. 5. Par 41-55: What about ebullition? You don't need saturated concentrations of methane to form bubbles (see Chanton and Dacey 1991and Chanton and Whiting 1995)Chanton, J.P. and Dacey, J.W.H., 1991. Effects of vegetation on methane flux, reservoirs, and carbon isotopic composition, In "Trace Gas Emissions by Plants", edited by TD Sharkey, EA Holland, and HA Mooney. Chanton, J.P. and Whiting, G.J., 1995. Trace gas exchange in freshwater and coastal marine environments: ebullition and transport by plants. Biogenic trace gases: measuring emissions from soil and water, pp.98-125.
We have removed mentions of ebullition in the introduction since this transport pathway is mostly relevant for CH4 and less for CO2 (more soluble). We nonetheless evaluate the potential role of ebullition in the regular losses of catotelm porewater CO2 in autumn in the discussion (Line 261-268).
6. Line 57-59: While Clymo did say this approx. a quarter century ago, more recent studies challenge this assertion. (see work by Paul Glaser). Now state that peatlands are traditionally viewed as two-layer systems (catotelm and acrotelm) and explain their respective roles in the peatland CO2 cycling (Line 28-44).
We now specify that the measurements are carried under low temporal sampling resolution (Line 53-54).
8. General intro: Although one of my papers was cited as supporting evidence, I wouldn't say that I believe deep porewater CO2 is "static". This statement has been removed (see reply to comment #1).
9. Lines 203-206: it seems contradictory to me to talk about a molecule of CO2 taking 1.5 years to reach the surface during one season. Although I think I know what the authors mean, it is just confusing. I think this would be better expressed as a comparison of velocities.
We now provide both speed and time estimate in this statement (Line 162-168). We consider that stating the transport rate in time units provides a more tangible unit for less familiar readers. Stating the time units is also informative to understand the rate of diffusion in relation with the sampling interval of porewater CO2 concentration (i.e., hourly measurements).
We thank the reviewer for providing this piece of information. We now provide more citations to this statement and include the study from Kolton et al 2019. However, our understanding of their results is that they report a peak in CO2 production at 5°C. This helps reinforce our argument that a slowdown in respiration is unlikely the main driver of the phenomenon observed in our data (Line 226-230).
12. Line 297-300: Something to note is that the water table in Sphagnum-dominated peatlands can be below the surficial sphagnum layer which may reduce wind-shear air-water exchange relative to the small ponds mentioned.
We now explicitly mention vegetation cover as a potential source of variability in the air-water gas exchange  3 This statement has been removed. We now provide more detailed calculations of the possible rate of porewater CO2 emission to the atmosphere (See comments from reviewer 2), based on the estimated air-water gas exchange coefficient (Methods Line 496-524), and porewater CO2 concentrations (Discussion, Line 284-304).
All measurements of porewater CO2 and CH4 partial pressure are converted to concentrations based on Henry's law (according to ambient temperature and pressure) (Methods Line 432-433 & 511-513).
The differences in solubility between CO2 and CH4 is also mentioned in the text (Discussion Line 263-265).
14. Line 439: should this be >2meters? Statement clarified  15. Line 440-442: I don't understand this. The deepest porewater (2m) was sampled once a month, but 2.5m was sampled hourly in some months? This part needs clarification.
Statement clarified  16. Paragraphs 522-552: I've done some diffusion modelling and I can't quite follow what they are doing here. For example, Is the mass transfer method (e.g., Happell et a., 1995) or the boundary layer method used to calculate exchange coefficient here? If the latter, the result is very sensitive to boundary layer thickness. How was that chosen? I think this section could use some clarification.

. Methane transfer across the water air interface in stagnant wooded swamps of Florida: Evaluation of mass transfer coefficients and isotropic fractionation. Limnology and Oceanography, 40(2), pp.290-298.
We have clarified this section of the methods (Line 496-524). We used the mass transfer method dictated by Fick's first law of diffusion (Eq. 1 in Happell et al. 1995 paper). Having continuous atmospheric CH4 flux measurements and the measurements of porewater CH4 concentration, we were able to estimate the air-water gas exchange coefficient (k600).
We are aware of the uncertainties of our k600 estimates (Line 294-296), but we still consider that these estimates are useful in determining the possible rate of increase in porewater CO2 emission to the atmosphere (Line 521-524), and the depth range of air-water gas exchange (Line 254-256). The estimated k600 values themselves are low and consistent with the environmental conditions and literature (Line 256-258).
We have recently deployed a CH4 sensor in the surface peat porewater near the eddy covariance tower and intend to use these measurements in a future study to model the air-water gas exchange coefficient in peat porewater with changing water table position and wind shear.
The manuscript presents evidence of a possible source of carbon dioxide losses from peatlands that has been ignored previously. The authors argue that turbulent diffusion has the potential to move deep porewater CO2 closer to the peatland surface, where it could be released by atmospheric emission and hydrological export. Their estimates of autumn loss of porewater CO2 storage for a Swedish peatland are comparable to the peatland's annual C sink. The authors challenge the perception that greenhouse gases in deep peat porewater are largely inert and raise the possibility of substantial errors in global estimates of peatland C sinks and GHG emission. The results are original and should be of interest to peatland researchers as well as to researchers with wider interests in the terrestrial carbon cycle and greenhouse gas emissions.
The key argument is supported by analysis of four years of near-continuous measurements of porewater CO2 and temperature at varying depths in a peatland in northern Sweden. The authors used a heat budget approach to estimate the turbulent diffusion coefficient for CO2. The temporal extent and resolution of porewater CO2 data are rare, if not unique to this study. The sampling strategy is valid for demonstrating recurring thermal instabilities and associated seasonal fluctuations in porewater CO2 concentrations and storage. Previous studies based on mineral solute profiles (e.g., Griffiths and Sebestyen 2016) have already demonstrated that deep porewater is dynamic rather than static. The unique contribution of this study is to demonstrate this phenomenon for dissolved CO2 and to propose that these seasonal porewater dynamics are explained by turbulent diffusion.
The section of the manuscript that considers the implications of turbulent diffusion for the peatland C budget is less convincing. The authors argue that rapid rise of deep porewater to the surface could release CO2 to the atmosphere and/or to runoff. They estimated the cumulative porewater CO2 loss through autumn as comparable to the peatland's annual net ecosystem exchange and an order of magnitude larger than its stream CO2 export, and the daily flux of deep porewater CO2 toward the surface during periods of instability as comparable to mean annual rate of ecosystem respiration. The difficulty is that the measurements of atmospheric net CO2 exchange and stream CO2 export do not show the pattern or magnitude of fluxes that would be expected. To account for this lack of evidence for rapid efflux of CO2, the authors argue that the method of partitioning NEE into GPP and Reco is flawed. Further, they argue that their peat depth profile may be unrepresentative of the whole catchment, explaining why autumn pulses of stream CO2 export were negligible compared to losses of porewater CO2 storage.
The possibility of substantial losses of deep porewater CO2 is tantalizing, but the evidence for its rapid efflux to the atmosphere or stream water is weak: based on the data available, the porewater CO2 budget is incomplete. It would be helpful to outline what measurements and experimental set-ups would be required to confirm and quantify C losses from deep porewater. For example, peat profile monitoring of hydraulic heads and the concentrations of a conservative solute would help to disentangle the contributions to vertical mass transport of advection, molecular diffusion and turbulent diffusion. Tracer studies would help to establish the fate of 'lost' porewater CO2 and, ideally, CH4. We agree with the reviewer and thank him/her for pointing out this issue. The previous assessment of the possible release of catotelm porewater CO2 towards the atmosphere or the stream outlet was indeed weak. We have revised these calculations and now provide a more thorough assessment of the potential increase in porewater CO2 emission to the atmosphere and hydrological export.

Reference
The previous manuscript used the time-series of porewater CO2 concentration to estimate apparent fluxes (i.e., based on daily changes at different depths). We consider that the resolution of our measurements across the depth profile was too low to generate reasonable estimates of the fluxes. For example, measurements at 1.5m were assumed to represent the full depth between 1m to 2m. These assumptions probably amplified the rate of porewater CO2 change per volumetric unit area. We removed those flux estimates, but nonetheless present an estimate of the continuous total porewater CO2 store, which we use to determine the magnitude of the regular losses in porewater CO2 store (Figure 6a, Methods Line 440-451).
We now estimate the average annual CO2 emission from the porewater to the atmosphere, based on k600 estimates (Method Line 493-524) and porewater gas concentrations, and determine the possible increase in atmospheric fluxes with a rise in porewater CO2 concentration in the surficial porewater (Discussion Line 284-304). We then compare this range in porewater CO2 emission with the time-series of ecosystem respiration (Line 296-297). This comparison reveals that ecosystem respiration rates in autumn generally exceeds this estimated range of porewater CO2 emission ( Figure 6c). This suggest that changes in porewater CO2 emission could be a significant contributing element to the annual ecosystem respiration. We also estimate the cumulative ecosystem respiration during the full period of weak thermal stability and demonstrate that it is similar to the periodic total loss of catotelm porewater CO2 store (Line 297-302, Figure 6a and c). We consider that these new calculations provide information on 1. The possible average annual porewater CO2 emission rate, 2. A reasonable range of increase in porewater CO2 emissions, 3. Indication that porewater CO2 emission, and their possible change over time, likely contribute to the measured ecosystem respiration at the site. We also no longer state that the GPP and ER partitioning method is flawed.
For the consideration of the possible increase in hydrological CO2 export, we now present the full timeseries of stream CO2 export (Figure 6b), and estimate the cumulative CO2 export during the periods of weak thermal stability. These results still confirm that the magnitude of the stream CO2 export per catchment area is about one order of magnitude lower than the loss of porewater CO2 in autumn. We nonetheless provide additional references to state that hydrological CO2 export may be greater at this location than over the rest of the catchment area, owing to the presence of a preferential flow layer at this site (2 to 2.5m deep) and the site's proximity to the stream outlet (i.e. flow convergence area) (Line 317-319).
The reference (Griffiths et al, 2016) is now presented in the discussion (Line 240-241).
This revised version of the manuscript also includes a paragraph stating the possible environmental condition at this site that could make it more prone to turbulence in peat porewater (Line 206-219). We formulate several recommendations for more detailed studies throughout the discussion e.g., 1. the possible interplay between increase in mass flow and turbulent diffusion in autumn (Line 202-204, 249-251), 2. The role of changing porewater CO2 store in the peatland C budget (Line 302-304, 322-323). We also highlight the latter element of uncertainty in the introduction (Line 65-66). We are currently working on several follow up studies and have instrumented a new peat depth profile near the eddy-covariance tower at Degerö Stormyr and at another nearby and newly-restored peatland.
Specific comments: Line 36 -reference needed for 'hydro-physically inactive'

Statement removed
Line 41 -Does 'CO2' refer to concentration or stock? Spelling: 'instantaneous' Statement revised. Spelling corrected Line 60 -'manual characterization with a low time resolution' -rephrase for clarity, e.g., periodic manual measurements?

Statement clarified
Lines 124-126 -Near-surface porewater CO2 concentrations will also be affected by rhizosphere processes, which change seasonally.
Rhizosphere processes are now mentioned (Line 73) Line 195 / Figure 5b -Given the relationship stated in the equation, it would make more sense for Kz and the trend lines to be plotted against 1/N2 rather than N2.
We indicate the direction of weak and strong thermal stability on the axis itself ( Figure 4b Overview This paper is a thought-provoking account of a high-quality dataset from a northern peatland that suggests that CO2 dissolved in deep (> 0.5-0.75 m depth) porewater can be lost from peatlands during the autumn. Previously, this store of CO2 has been assumed to be largely static or having a slow turnover. In a mostly well-written paper, the authors suggest that their dissolved CO2 data can be explained by the process of turbulent diffusion, which can lead to rates of dissolved CO2 transport that are orders of magnitude greater than molecular transport. In a careful analysis, the authors also consider the importance of the posited process compared to other processes in the carbon cycle in peatlands. Overall, I enjoyed reading the paper and I believe there is information in it that is very worthy of publication in Nature Communications. However, I struggled to understand parts of the manuscript and I believe the authors could do a better job of explaining their proposed mechanism. Below, I explain how I think the paper could be improved. I have also marked up the pdfs of the manuscript and the SI, and the editor and authors are referred to these for other comments and concerns that I recommend are addressed.
More than one type of turbulent diffusion?
Groundwater scientists usually consider there to be two ways by which solute mixing may occur: molecular diffusion and mechanical dispersion, the latter being analogous to diffusion but occurring in laminar flows due to the 'splitting' of flow through a porous medium as flowing water 'samples' the wide range of pore sizes. Turbulence can sometimes occur in porous media in, for example, biogenic macropores such as wormholes, larger voids between soil peds, and joints in bedrock. Turbulent transfer of a range of properties is also known to occur in surface waters, and the transfer has been found to follow mathematically the diffusion equation; hence, the use of the term 'turbulent diffusion'. The latter occurs when there is rapid mass transfer of water. The authors, however, seem to be suggesting a type of turbulent transfer that does not involve mass flow of water. From the text and Figure 6 they propose that pore-scale turbulent eddies form that help transfer dissolved CO2 along a concentration gradient. They also note that this turbulence causes a transfer of CO2 that is orders of magnitude greater than would occur via molecular random walks.
In principle, I find this suggestion plausible, but what I find confusing is the authors' explanation, or lack of explanation, of the mechanisms involved. In short: what causes the turbulence? The authors are clear in suggesting mass flow is not responsible (e.g., on lines 226-7), and they warn against treating the pore-water domain in the same way as lake water where turbulent diffusion has also been observed. Perhaps it is my ignorance showing here, and perhaps other scientists will clearly understand what the authors mean. However, I suspect many target readers of the paper will be like me and not know how turbulent diffusion can occur without mass flow. I guess part of my problem here is that I struggle to see how turbulent mixing at the scale posited by the authors can occur in the pores found in the deeper peat at their site. They suggest that turbulent eddies in the peat pores might form at scales of mm to cm but I don't see how that is possible because most pores in the deep peat will be of mm scale or much less (based on the hydraulic conductivity data presented in the SI). Notably, on lines 227-235, the authors themselves seem somewhat uncertain about how turbulent diffusion can occur in peat in the absence of mass flow.
We thank the reviewer for identifying those aspects which needed further clarification in the manuscript. Indeed, a high level of clarity on the mechanism of turbulent diffusion is essential in this paper. We hope that our changes have made our explanations clearer:

Source of turbulence
Turbulence consists of small-scale random fluid motion (eddies) propagating in the peat porewater (described as Kz by Osborn et al, 1980), which is different to directional flow of water through the peat pores (determined by Ksat, Figure S2). Vertical gas diffusion in water occurs at the molecular level (slow and constant) when the water is completely still and contains no kinetic energy (generally valid in groundwater). However, the high porosity of the peat porewater, its exposure to wind shear on the peatland surface and flow through the peat pores provide a source of kinetic energy in the peat porewater. Hence, the porewater is not still, and contains a very small level of kinetic energy (propagating as turbulence), which increases vertical diffusion relative to the molecular rate (i.e., turbulent diffusion, variable in time and space). Kinetic energy input provides the source of turbulence, and changes in porewater temperature stratification determines the degree of suppression or propagation of this kinetic energy in the form of turbulence (i.e., small scale eddies) in the catotelm porewater. Weak thermal stratification (weak stability) allows turbulence to propagate, while a strong thermal stratification (strong stability) suppresses turbulence. This shift in turbulence propagation can occur even under constant and low kinetic energy input.
We improved the discussion (Line 127-219) and method (Line 453-490) sections and the schematic ( Figure 5) to further clarify the mechanism of turbulent diffusion. We provide further clarification of the model of turbulent diffusion, described by Osborn (equation 1). We explain the interplay between kinetic energy and changes in porewater thermal stability (equation 1) and discuss the estimated levels in kinetic energy in the porewater (Line 150-158). Figure 4b now displays the estimated kinetic energy levels (oblique lines) in different porewater depths. This reveals that kinetic energy in the peat porewater is very low (on average one order of magnitude lower than in small and sheltered lakes Line 154-156), but also considerably higher in the surficial porewater than deeper peat horizons. We also state the possible source of this kinetic energy (Line 157-158) (i.e., wind shear near the peat surface and lateral flow through the peat pores) and also point to the possible role of ebullition as another source of kinetic energy in the catotelm porewater (Line 265-266, see next comment).
We did not intend to suggest that there was more than one type of turbulent diffusion, but rather that diffusion can occur at other rates than strictly molecular (i.e turbulent). We emphasize the differences between molecular and turbulent diffusion (e.g., turbulent diffusion is defined as vertical diffusive transport that exceeds the molecular rate (10 -7 m 2 s -1 for heat transfer) (Line 475-476). Unlike molecular diffusion, which is constant and slow, the rate turbulent diffusion can vary widely in time (Line 130-132). We also highlight that the relationship between our estimates of Kapp and N 2 , to clarify the link between turbulent diffusion and the weakening thermal stratification (Figure 4b, Line 160-170).
The schematic ( Figure 5) now illustrates the models of all three diffusion rates a) molecular b) turbulent under strong thermal stability, c) turbulent under weak thermal stability. This makes the differences between molecular and turbulent diffusion clearer. The turbulence in porewater (drawn in black arrows) is now better illustrated in the form of "random fluid motion". The depth profile now indicates the ksat profile (grey arrows), interpreted from the bulk density measurements at the site. The turbulent diffusion is indeed separate from directional flow of water (i.e., mechanical dispersion (or mass flow) via convection or advection) ( Figure 5).

Scale of eddies
In the previous version of this manuscript, we mistakenly stated that the scale of the eddies could be of mm to cm scale, while these are in fact the sizes of large eddies measured in lakes. The estimated Kapp, and kinetic energy in the peat porewater (equation 1) are on average one order of magnitude lower than in small sheltered lakes (Line 154-155). We have rectified this statement, now stating the scale of the eddies is lower than mm scale, which is more consistent with the pore size in the catotelm and the literature (Line 187-189).

Properties of the studied site
We added a paragraph in the discussion stating the possible properties of our study site that could make this location more prone to turbulence. This includes 1. Relatively high peat porosity, 2. Presence of a deep preferential flow path and in close proximity to the stream (flow convergence area), which might increase specific CO2 export. 3. Peatland exposed to strong wind since it is located at a topographic high point (Line 206-219).

Ebullition and measurement of dissolved CO2
I wonder too whether bubble formation and transport within the peat might also offer a part explanation of the authors' findings. Biogenic gas bubbles tend to accumulate in peat during the growing season and reach a peak unit volume by late summer and early autumn. These bubbles contain CO2 as well as CH4 and their movement upwards through the peat profile may also facilitate CO2 transfer. Their movement through, and loss from, the peat profile will also cause mixing of pore water that might give rise to turbulent diffusive transfer of dissolved CO2. I should add that sudden episode of gas bubble loss from peat are not always easily detected or measured by eddy correlation. The flux tower used by the authors was also somewhat distant from the monitored peat profile so may has missed ebullition events near the sensors.
We agree with the reviewer that ebullition could play a role in the porewater CO2 dynamics. Yet due to the relatively high solubility of CO2, most of the CO2 pool is in dissolved rather than free-phase. Diffusion should therefore be the main transport pathway for CO2, despite a share of the gases possibly transported in bubbles. We now discuss the potential role of ebullition in the porewater CO2 dynamics observed in our data (Line 261-268). We also indicate that the ebullition could provide an additional source of kinetic energy and turbulence in the catotelm porewater (Line 266-268). However, we still ascribe the increase in turbulent diffusion with the weakening of porewater thermal stability as the main process responsible for the sudden losses in catotelm porewater CO2 store in autumn (Line 268-270).
I have a slight concern about the measurement of dissolved CO2. What happens to the CO2 sensors if bubbles are in contact with the sensor membrane? Is it possible that the sensors measure a combination of dissolved and free-phase CO2?
The sensor could measure a combination of dissolved and free-phase CO2 (e.g., if bubbles were stuck on the membrane). But as stated previously, the majority of the CO2 pool should be in dissolved phase.
Also, if bungs were fitted to the tops of the sampling tubes, how did flow of water into and out of the tubes occur? Such tubes need to be vented to allow water ingress and egress.
The tubes have thin slits screening for specific depths (Line 406-408). Flow in and out of the tubes should be proportional to peat hydraulic conductivity at those specific screened depths.
The tubes were "vented" (bung removed) in Spring 2015 and 2016, while the membrane of the sensor was replaced (now stated Line 413-415). We noticed large decreases in CO2 concentration following those periods where the tubes were left open for several hours/days. The tubes probably acted as chimney for porewater gas evasion to the atmosphere. The CO2 concentration returned to their initial levels several days/weeks after the tubes the closed, which resulted in a large gap in a time-series, specifically during spring. We subsequently refrained from leaving the tubes open for extended periods of time. In the following year (2017), the tubes remained closed in order not to affect conditions during spring thaw.
Tubes were also opened for a few minutes/hours periods when sensors had to be retrieved for winter (every November, near the surface only) or when sensor damaged forces us to redistribute the sensors at different depths follow sensor malfunction. We did not notice large changes in CO2 concentrations following those shorter manipulations.

Recommendation
Overall, I recommend the paper is revised and reviewed again. I think the authors need to explain their proposed mechanism more clearly, so that wider readership can understand it. Currently, I have some doubts about what the authors are proposing, and I'd like to be more convinced that their mechanism is sound.
We thank the reviewer for his constructive comments and hope that the revised version of the manuscript will adequately answer his concerns.
Please note that I operate a policy of 'open reviewing'. My comments above, and on the marked-up manuscript and SI, are for the editor and the authors and I would like my identity to be revealed to the authors.

Comment by lines:
Line 33. Why latent?

Statement removed
Line 35 Avoid using 'this' in this way (i.e., say what 'this' refers to). 'This lack of oxygen ...'.

Statement corrected
Line 37: Unclear in two ways. What processes are implied by 'hydro-physically'? 'lower depths' suggests shallower (nearer the surface) peat. Don't you mean greater depths?. By whom? I'm not sure there is such a general perception.
We removed this statement. Peat hydrology is indeed dynamic. We rather claim that the processes driving the peatland CO2 cycling (i.e., atmospheric exchange and hydrological export) operate mostly within the acrotelm (Line 32-44).
Line 52: I think there is a misunderstanding here. Horizontal flow can only occur if there is vertical recharge. The latter then becomes the rate limiter. Vertical and horizontal K cannot be separated from each other when calculating flow through an aquifer, which is not the same as uni-directional flow through a permeameter sample of peat (for example).

Statement revised
Line 61: Just say 'store' here.
"Storage" changed to "store" here and throughout the text Clearer definition now provided (Line 90-107). We also changed the terminology from thermal stability and instability to strong/weak thermal stability.

Line 89: Okay, so what mechanism(s) is(are) involved?
We now provide more clarification of the mechanism of turbulent diffusion and the interplay with thermal stability here and throughout the text (see response to first comment) Line 95: Rather vague wording.
Exact dates with length of the period are now specified (Line 84) Line 97: Why is it a thermal instability? What is meant by 'instability' here?
We explain the link between strong/weak thermal stratification and thermal stability (Line 91-94). We specify thermal stability because these calculations are based on temperature measurements, but stability (buoyancy) can also be affected by solute concentration. We also refer to strong/weak thermal stability instead of thermal stability and instability.
Line 116: Just the CO2 or the porewater too?
We hope our improvement to the text have helped clarify this aspect. Since we refer to diffusion and not large-scale directional fluid motion (only small-scale random fluid motion (turbulence))(Line 138-139), it is just the CO2 that diffuses upwards due to its gradient (Line 128-140).
Line 127: -7d? now clarified 7 days moving window (Caption Figure 2) Line 134: But if hydraulic conductivity is low, density-driven flow of water will also be low.
We do no invoke large-scale density-driven flow of water (Line 198-201, Figure S5), only small-scale random fluid motion in the peat porewater (Line 138-139).
Line 137: How so? I don't follow.
We clarify the possible influence of ice-cover on porewater turbulence (Line 177-179) Line 158: Minus signs missing here.
All depths are now referred to without minus sign Line 161: There is a lot of information in each plot. Would it be better to select just a few key temperature profiles and use these to illustrate the point being made?
The plots (Figure 3) now show key examples, which we agree, makes the figure much clearer.
Line 186: I agree. However, it would be good to explain a little more about turbulent diffusion here -please see my report. The paper could be made more accessible to those, like me, who were unaware of the proposed mechanism.
We added more explanations in the text as well as improved the schematic Figure 5 Line 202: Four orders of magnitude? We reworked this section of the text to explain that the increase in vertical diffusion in autumn may not be accompanied by an increase in convection, as referred to in lakes with turbulent mixing (Line 186-204).
Line 241: But is there not threshold too at about 4 deg. C below which respiration shuts down?
We now provide a reference stating that respiration can even peak at ~5°C (Line 229) Line 244-245: Unless they cause ebullition. Also, in what way are the deeper waters confined? If the peatland is a water-table aquifer then it is unconfined. Statement revised. Figure 6: This is a nicely-designed figure, but I think it is misleading to suggest lateral export at 2 m depth is the same as that near the surface. Perhaps I have misunderstood something here. Also, shouldn't the CO2 emission be larger in the right-hand figure?
Figure edited (see response to first comment) Line 289: The surface porewater is not hydrologically conductive; the peat is. Sentence needs rewording.

Statement corrected
Line 326: Actually, it could occur from a range of depths. There is a lot of work that shows CH4 production at shallow depths too; e.g., just below the water table where there is abundant substrate and also sufficient anoxia. Corrected Line 397: I agree this is very low. It is not typical of peatlands more widely. Was the study site therefore quite unusual?
We now present a section discussion the particularities of our studied location in connection with the mechanism of turbulent diffusion (Line 206-219).
Line 407: Name / proper noun? In which case start the words with capitals? Corrected Line 431: 'bungs', But the bungs would also have affected water flow into and out of the tubes, so how can you be sure that the water in the tubes was in equilibrium with the pore water?
We assume that the speed of water inflow and outflow through the tubes is proportional to the hydraulic conductivity at the porewater layer screened by the tube. We measured considerable seasonality in porewater solute concentration and isotopic ratio (Campeau et al 2017 and2018), and water temperature (Figure 1b), which indicates continuous water recharge in the tubes. We agree with the reviewer that there is always a level of uncertainty when undertaking measurement in groundwater tubes. However, deploying the sensors (with a fragile membrane) would not have been possible without those groundwater tubes. We recently lauched a follow up study where we have instrumented another peat depth profile in close proximity with the eddy-covariance tower. For this new instrumentation, we invested in new and more sturdy sensors (eosense GP), which we could deploy directly in the peat, without requiring the installation of groundwater tubes.
Line 467: 'to account for' Corrected Line 476: m-2? Corrected Line 531: Okay, but was there any evidence of enhanced ebullition during the autumn period? Not to our knowledge, but a more detailed examination of CH4 ebullition at the site is planned for the future. Figure S2: The figure caption says m below ground surface. Also, there is the potential for confusion with the use of negatives. A negative depth (below ground) means something is above the surface. I suggest being clearer in the axis label (something like 'water table position relative to the ground surface' would be okay). Also, why does the water table rise above the surface during the ice/snow covered period. Is the peat at that time actually a confined aquifer when the concept of water table does not apply?
Depth reporting are adjusted here and throughout the text.
The positive water table measurements in winter are caused by the accumulated snow and ice cover putting additional pressure on the pressure transducer. These measurements are highly uncertain, so they have been removed. These data are now presented in Figure 1d. Figure S3: What measurement method was used?
The methods are detailed in [Nijp et al., 2019]. We nonetheless specify that they were inferred based on piezometer measurements of water table depth.
Andy Baird, Chair of Wetland Science University of Leeds, UK; 13th April 2021.

•
Removed monthly measurements of porewater DOC concentration from Figure S5, since they did not convey more information than the water stable isotope measurements. These data are already presented in (Campeau et al, 2017 and2018).

•
Changes the heatmap of porewater temperature for the time-series, to show more clearly periods of temperature equilibration, and for comparative use with other studies in the future • Main text includes the water table position time-series (Figure 1d) since this is usually an important variable in peatland studies.
The authors have done a good job of responding to all of the points raised by the reviewers. I do not have any additional significant concerns. I would suggest some minor revisions to the figures to make them more legible. For example Figure 1 I would recommend that the text of the figure legends be a larger font size.
The same for the text on the panels in Figure 3; they were difficult to read at 100%. For figure 3, I might also suggest making the two series a different symbol shape as well as color to further distinguish them. Finally there appears to be a typo on line 530 should be "atmosphere".
Reviewer #2 (Remarks to the Author): General comments: Turbulent diffusion is a new idea for peatland scientists, and the evidence and explanation presented in this manuscript clearly demonstrate its potential implications for peatland carbon cycling. The revisions improve the manuscript by clarifying the proposed mechanism of turbulent diffusion and by providing a more cautious appraisal of the implications for CO2 emission and water-borne export. Overview I reviewed the original submission of this paper and have been asked to comment on whether the revision takes account of the concerns raised in my original report. I also read the reports of the other two reviewers, and the authors' explanations of how they responded to the first round of reviews. Overall, I think this revision is a considerable improvement on the original. In particular, it is much clearer in its explanation of turbulent diffusion and the mechanisms involved. However, I think some further changes will make it more convincing. Upon re-reading the work, I have also identified a possible problem with the data set. My more substantive concerns are articulated below. I have also added comments to a pdf of the paper to which the editor and authors are referred.
CO2 production in autumn and winter? I apologise for not discussing this issue in my first review, but, looking at the data again, I am struck by the rapid increases in porewater [CO2] in the late autumn and early winter. The authors interpret the rapid latesummer and early-autumn declines in porewater [CO2] as a substantial loss of CO2 from their monitored peat profile, but where does the CO2 come from for the rise that follows? What plausible mechanisms can be invoked? CO2 production will occur slowly in the catotelm because of anoxia and will also be slow in the acrotelm as temperatures fall. This apparent discrepancy needs to be resolved if we are to believe the data. The authors did not measure 3-D patterns of porewater [CO2], and it is possible that CO2 is moving laterally in a complex way through the peat profile, and such movement may help explain the data. In other words, the 1-D patterns observed by the authors may not apply laterally across the peatland.
Piezometer tubes In my original review I noted: "Also, if bungs were fitted to the tops of the sampling tubes, how did flow of water into and out of the tubes occur? Such tubes need to be vented to allow water ingress and egress." In their rebuttal, the authors note "The tubes have thin slits screening for specific depths (Line 406-408). Flow in and out of the tubes should be proportional to peat hydraulic conductivity at those specific screened depths." In the revised paper they also note: "The top of the wells was sealed with thick rubber bongs [sic] to prevent atmospheric gas exchange." If a well is screened below the lowest position of the water table and is sealed at the top of the lining tube, it will not function properly as a piezometer, regardless of the hydraulic conductivity and hydraulic gradients in the soil. Air trapped in the well tube will act to stop water ingress and egress. If, for example, porewater pressures rise outside the screened section, water will flow into the well to equilibrate to the new pressure. However, as water begins to flow into the piezometer it will start to compress the air, which becomes pressurised, thus preventing more water entering the tube. A similar argument applies when porewater pressures in the soil around the piezometer fall. This problem is well known among hydrogeologists, but it is unclear whether it produced artefacts or errors in the authors' porewater [CO2] data. I think we can be confident that water flow into and out of the tubes was inhibited and, therefore, that the water around the CO2 sensors will have been 'old' water. However, I suspect the distance between the sensor membrane and peat outside the piezometer (of the order of 8 mm allowing for tube wall thickness) was probably short enough to mean dissolved CO2 by the membrane was in quasi equilibrium with porewater CO2 in the peat (equilibration occurring via molecular and turbulent diffusion).
Thermally-driven convective flow? On line 480, the authors correctly note that the hydraulic conductivity of the catotelm is almost certainly too low for thermally-driven convective flow of porewater to occur (see, for example, https://www.pnas.org/content/100/25/14937.short). However, in saying this the authors appear to contradict what they say on lines 207-208. I recommend they rewrite what is on those lines so that it is consistent with what they say later in the paper.

Soil respiration at 3-5C
On lines 231-235, the authors note: "The temperature dependency of peat soil respiration is mostly linear and thus inconsistent with a rapid shutdown at 3-5 °C 39, 40, 41, which would be necessary to explain the phenomenon in autumn revealed by our data. One study even reported peak CO2 production in peat porewater around 5°C 42, which would be opposite to the losses in catotelm porewater CO2 store observed each autumn in this peatland." I'm still not convinced by what is written here -it reads a little like special pleading. There is a huge body of knowledge that shows that ecosystem respiration (ER) in peats is low or shuts down at ~4 C -see, for example, the very large flux chamber literature (the authors' own ER data also show this). I know ER includes aerobic decay in the acrotelm and plant respiration, but it also includes anaerobic microbial respiration in the catotelm. I think the discussion here could be more nuanced/balanced.

Units
In several places units don't seem to be given correctly or are not given at all. For example, at times I was confused about whether the diffusive flux or the diffusion coefficient was meant. Elsewhere, units for kinetic energy dissipation are seemingly used for kinetic energy.
I continue to operate a policy of 'open reviewing'. My comments above, and on the marked-up manuscript, are for the editor and the authors and I would like my identity to be revealed to the authors.
Andy Baird, Chair of Wetland Science University of Leeds, UK; 5th August 2021.

3
The authors have done a good job of responding to all of the points raised by the reviewers. I do not have any 4 additional significant concerns. I would suggest some minor revisions to the figures to make them more legible.

5
We thank the reviewer and are happy that he/she is satisfied with our revision of the manuscript 6 7 For example Figure 1 I would recommend that the text of the figure legends be a larger font size. The same for 8 the text on the panels in Figure 3; they were difficult to read at 100%.

9
Font size is increased in both figures 10 11 For figure 3, I might also suggest making the two series a different symbol shape as well as color to further 12 distinguish them.

13
Symbol for temperature data is now different from the CO 2 concentration data

15
Finally there appears to be a typo on line 530 should be "atmosphere".
General comments: Turbulent diffusion is a new idea for peatland scientists, and the evidence and explanation 20 presented in this manuscript clearly demonstrate its potential implications for peatland carbon cycling. The 21 revisions improve the manuscript by clarifying the proposed mechanism of turbulent diffusion and by providing a 22 more cautious appraisal of the implications for CO2 emission and water-borne export.

23
We thank the reviewer and are happy that he/she is satisfied with our revision of the manuscript

25
Specific comments: 26 Lines 43-44 -'contribution to the peatland CO2 cycling' is a bit vague, and I'm not sure what it would mean for 27 this contribution to be 'low'

28
Sentence now rephrased (Line42-43): "As long as CO 2 remains confined in the catotelm, its role in the 29 peatland CO 2 cycling is negligible"

34
Lines 54-55 -This paragraph needs to end on a clearer statement of the message, e.g., assumption of slow 35 dynamics and sampling at coarse temporal resolution (state timescales?) may miss seasonal dynamics driven by 36 process(es) other than diffusive transport?

37
New sentence added at the end of the paragraph to more explicitly define the knowledge gap (Line 54-

97
Overview 98 I reviewed the original submission of this paper and have been asked to comment on whether the revision takes 99 account of the concerns raised in my original report. I also read the reports of the other two reviewers, and the 100 authors' explanations of how they responded to the first round of reviews. Overall, I think this revision is a 101 considerable improvement on the original. In particular, it is much clearer in its explanation of turbulent diffusion 102 and the mechanisms involved. However, I think some further changes will make it more convincing. Upon re-103 reading the work, I have also identified a possible problem with the data set. My more substantive concerns are 104 articulated below. I have also added comments to a pdf of the paper to which the editor and authors are 105 referred.

106
We thank the reviewer for his constructive feedback. A detailed response to his concerns follow: where does the CO2 come from for the rise that follows? What plausible mechanisms can be invoked? CO2 113 production will occur slowly in the catotelm because of anoxia and will also be slow in the acrotelm as 114 temperatures fall. This apparent discrepancy needs to be resolved if we are to believe the data. The authors did 115 not measure 3-D patterns of porewater [CO2], and it is possible that CO2 is moving laterally in a complex way 116 through the peat profile, and such movement may help explain the data. In other words, the 1-D patterns 117 observed by the authors may not apply laterally across the peatland.

119
We agree with the reviewer that the rate of recovery in early winter needed more formal analysis (now 120 presented on . We now provide further information on the recovery of porewater CO 2 in 121 early winter (only measured at 0.75 m and 1.5 m depth). Figure Table S1 in the supplementary files.

129
There were, however, two years where the recovery in porewater CO 2 at 0.75 m was faster (i.e. 2014 and 130 2017, Figure 1a of response file & Figure S1a of supplementary file). The rate of recovery was above 0.2 131 mg C g -1 d -1 for 5 days in 2014 and 2 days in 2017, with a maximum of 1.2 mg C g -1 d -1 (Figure 1 in 132 response letter). We consider that those high rates of CO 2 increase are likely the result of additional 133 transport processes supplementing biogenic production. In those same two years (unlike the other years) 134 the porewater CO 2 concentration at 1.5 m depth decreased sharply during autumn, indicating that the 135 destabilisation of porewater CO 2 store reached deeper into the peat profile in those years. A likely 136 explanation is that the more severe destabilisation of the catotelm porewater CO 2 store on those years was caused by higher turbulence or a prolonged period of thermal instability, which could allow CO 2 from 138 deeper peat horizons to diffuse upward, or laterally from other areas of the peatland, and assist in the

155
We also added a few sentences at the beginning of the results and discussion section (line 73-77) to 156 remind the readers that the concentration timeseries reflect the net balance between CO 2 input (transport 157 and production) and outputs (mostly transport). We consider that it was useful to set the basis for further 158 discussion on CO 2 loss and recovery.

160 2. Piezometer tubes 161
In my original review I noted: "Also, if bungs were fitted to the tops of the sampling tubes, how did flow of water 162 into and out of the tubes occur? Such tubes need to be vented to allow water ingress and egress." In their 163 rebuttal, the authors note "The tubes have thin slits screening for specific depths (Line 406-408). Flow in and out 164 of the tubes should be proportional to peat hydraulic conductivity at those specific screened depths." In the 165 revised paper they also note: "The top of the wells was sealed with thick rubber bongs [sic] to prevent 166 atmospheric gas exchange."

168
If a well is screened below the lowest position of the water table and is sealed at the top of the lining tube, it will 169 not function properly as a piezometer, regardless of the hydraulic conductivity and hydraulic gradients in the soil.

170
Air trapped in the well tube will act to stop water ingress and egress. If, for example, porewater pressures rise 171 outside the screened section, water will flow into the well to equilibrate to the new pressure. However, as water 172 begins to flow into the piezometer it will start to compress the air, which becomes pressurised, thus preventing 173 more water entering the tube. A similar argument applies when porewater pressures in the soil around the 174 piezometer fall. This problem is well known among hydrogeologists, but it is unclear whether it produced 175 artefacts or errors in the authors' porewater [CO2] data. I think we can be confident that water flow into and out 176 of the tubes was inhibited and, therefore, that the water around the CO2 sensors will have 177 been 'old' water. However, I suspect the distance between the sensor membrane and peat outside the 178 piezometer (of the order of 8 mm allowing for tube wall thickness) was probably short enough to mean dissolved 179 CO2 by the membrane was in quasi equilibrium with porewater CO2 in the peat (equilibration occurring via 180 molecular and turbulent diffusion).

182
This is an important point, and we agree with the reviewer that we could have described it better.

183
However, we argue that this impact on our data and analysis was minor. The major reason is, 184 as also pointed out by the reviewer, that the very short distance (6.5 mm) between the sensor membrane 185 and the external porewater allowed for continuous and rapid diffusive gas exchange between the 186 surrounding peat porewater and water inside the tube, despite a possible effect on water ingress and 187 egress. Our general findings of the importance of turbulent diffusion in peat porewater are supported by porewater gas concentration and temperature measurements, which would equilibrate by diffusion 189 between the outside and inside of the tubes, regardless of possible reduction in water transport.

191
Furthermore, the amplitude in groundwater table fluctuation across the year at our study site is very low.

192
The lowest water table depth is 19 cm relative to ground surface over the four years of observation. Under 3. Thermally-driven convective flow? 204 On line 480, the authors correctly note that the hydraulic conductivity of the catotelm is almost certainly too low 205 for thermally-driven convective flow of porewater to occur (see, for 206 example, https://www.pnas.org/content/100/25/14937.short). However, in saying this the authors appear to 207 contradict what they say on lines 207-208. I recommend they rewrite what is on those lines so that it is consistent 208 with what they say later in the paper.

210
We thank the reviewer for pointing out the apparent contradiction, and for providing this reference. We

216
On lines 231-235, the authors note: "The temperature dependency of peat soil respiration is mostly linear and 217 thus inconsistent with a rapid shutdown at 3-5 °C 39, 40, 41, which would be necessary to explain the 218 phenomenon in autumn revealed by our data. One study even reported peak CO2 production in peat porewater 219 around 5°C 42, which would be opposite to the losses in catotelm porewater CO2 store observed each autumn in 220 this peatland." I'm still not convinced by what is written here -it reads a little like special pleading. There is a 221 huge body of knowledge that shows that ecosystem respiration (ER) in peats is low or shuts down at ~4° C -see,

222
for example, the very large flux chamber literature (the authors' own ER data also show this). I know ER includes 223 aerobic decay in the acrotelm and plant respiration, but it also includes anaerobic microbial respiration in the 224 catotelm. I think the discussion here could be more nuanced/balanced.

225
We agree with the reviewer that this section needed to be better articulated. We have reworked this 226 section and provided further support for our claim that dropping porewater CO 2 production alone cannot 227 explain the losses in porewater CO 2 present in our data (Line 254-267). This section includes the     As noted by the reviewer, the contribution from catotelm heterotrophic respiration should be smaller in 242 comparison with the autotrophic (above and belowground) and the acrotelm (mostly aerobic) 243 heterotrophic respiration. The rapid shutdown in ER around average daily temperature of ~4°C is often 244 attributed to plant senescence following the first episode of soil frost at night. For these reasons, the temperature sensitivities for peatland ER is poorly comparable with catotelm CO 2 production dynamics, 246 which are the focus of this paper.

Units 249
In several places units don't seem to be given correctly or are not given at all. For example, at times I was 250 confused about whether the diffusive flux or the diffusion coefficient was meant. Elsewhere, units for kinetic 251 energy dissipation are seemingly used for kinetic energy.

258
Line by Line Comments:

260
Line 15: This is a specialist term used and understood only by peatland scientists. Perhaps define it here for more 261 general readers.

262
We agree with the reviewer that the term catotelm is not appropriate in the abstract. Due to the limited 263 word count, it was difficult to fit in the definition of catotelm. Hence, we simply removed the term and 264 kept our explanation to "deep porewater" which is the term that is also used in the title

266
Line 16: Do you actually mean 'temporally' (over time)? 'temporarily' means for a short period of time. Perhaps 267 just delete because 'stable' alone is clear.

299
See response to comment 1 above. We added a section providing a more detailed interpretation of the 300 CO 2 recovery during winter , We also present the detailed time series of porewater CO 2 301 concentration and daily changes in concentration at 0.75 m and 1.5 m throughout autumn and early-302 winter, indicating that most periods of recovery can explained by local biogenic CO 2 production (Figure 303 S1). Table S1 also details the rate of CO 2 recovery in early-winter at both depths for each individual year.

304
With respect to the annual temperature variation, maxima occur between August and October at 0.75m 305 and 1.5m depth. With respect to the kinetic constraints on peat decomposition, the highest annual 306 production rates possibly occur during autumn.