the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 06 Mar 2025
Amazon rainforest ecosystem exchange of CO2 and H2O through turbulent understory ejections
Abstract. We investigate the role of short-term variability on the mean ecosystem exchange of carbon dioxide and water vapor. Specifically, we focus on quantifying how the intermittent turbulent exchange at the forest-atmosphere interphase – characterized by sweeps, ejections and outward/inward interactions – contributes to the mean exchange. To this end we analyze observations of high-resolution (isotopic) flux measurements taken at 25 m above the forest canopy at the Amazon Tall Tower Observatory (ATTO) during the dry season. We identify short-term turbulent eddies that eject carbon dioxide and water vapor from the understory (0–15 m) into the atmosphere. The H2O ejected from the understory is shown to be depleted in deuterium (2H) by 10 ‰ compared to H2O originating from the top canopy. We show that this matches the depleted water vapor isotopic compositions found in understory leaf and soil samples. The diurnal cycle of the net ecosystem exchange (NEE) of CO2 is presented as a function of the sweeping and ejection motions and understory flux contributions. Understory contributions average 1.4 % of NEE, but reach up to 20 %. In exploring the connection between intermittent canopy turbulence and cloud passages, we found a weak but coherent temporal relationship (r = 0.027) between cloud passages and ejections, without a predominant influence of large clouds. These findings deepen our understanding of the gas exchange of the Amazon rainforest, which is urgent for predicting and possibly preventing the regions transition from a carbon sink to a source.
- Preprint
(4753 KB) - Metadata XML
- BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on egusphere-2025-452', Peter A. Taylor, 14 May 2025
- AC1: 'Reply on RC1', Robbert Moonen, 10 Jul 2025
-
RC2: 'Comment on egusphere-2025-452', David Bowling, 28 May 2025
This is an excellent and novel paper focused on using high-frequency turbulence observations to evaluate the dynamic nature of rainforest canopy venting events. It is quite appropriate for the journal and generally of high quality. The datasets are quite valuable and unique and are likely to be used by others. I did not download the data but did verify that the authors have posted the data publicly already.
It appears that the authors did not extract xylem water for isotopic analysis (no data are shown). This was an unfortunate oversight for such an isotope-intensive project. For this reason, your claims about knowing the isotopic signature of understory transpiration are rather weak. I highly recommend tuning up the modern Craig-Gordon model (as described by Cernusak et al. 2016) to calculate transpiration signatures based on the data in Fig 3B.
The introduction is well written and concise, with a clear hypothesis. I have many mostly minor comments that should be addressed before publication. However, this paper definitely should be published! Nice work!
It would be great to see a version of Fig 2 for sensible and latent heat fluxes also, and maybe add typical profiles during midday of the various scalars.
Abstract: this statement is incorrect "We show that this matches the depleted water vapor isotopic compositions found in understory leaf and soil samples". Understory leaf water was about +8 permil dD, and soil water was -15 to -20 permil (Fig 3B). All of these values are more enriched than the background atmosphere and the estimated values of source composition (Fig 3A). Where is the match?
Minor comments:
3: forest atmosphere interface (not interphase)
12: weak but coherent (r=0.027) - unclear meaning upon reading abstract
14-15: how can a deeper understanding of gas exchange prevent the transition from C sink to source?
48: 20 Hz sampling rate here, but Fig 1 caption says 10 Hz
60: improved and improvement redundant
74: other usage of the hyperbolic isolation function for isotopic exchange can be found here: Bowling et al (1999) JGR 104:9121, and (2003) AFM 116:159
76: for isolation OF wind sweeps and ejections
79: 2 sigma represents standard deviation of the residuals from the ODR fit? - more mathematical detail needed here so others can reproduce this
80: ejections are Q1 not Q2 according to your notation in Fig 1 (your notation is consistent with Thomas et al. (2008) and those events were identified by them as subcanopy-venting updrafts - please be clear that the labeling of quadrants is dependent on the variables plotted - the Shaw ejections in Q2 (your Fig A5) are not the same meaning as Q2 in your Fig 1
69-72: more detail would be helpful here - it seems you are splitting the total turbulent flux based on the quadrants, and for each quadrant taking a Reynolds average of the instantaneous product of w and scalars (or w and u)? so total turbulent flux = sum of turbulent fluxes in each quadrant?
80-83: more detail needed here - show us the equations that define how you calculate the bulk and ejection fluxes
87: are those 36 consecutive points or 36 from anywhere within the 7200?
88: one-sided uncertainties is vague, and more detail would be helpful on why "unrealistic source compositions" occurred
95: incomplete sentence
108-112: this is too vague to understand - this looks like a time-lagged cross correlation between "understory ejection intensity" and "cloud intensity"??? what are "two time series" and why did you not look for the time of maximum cross-correlation? - it seems that Fig 4 D should be labeled to be consistent with these variables (is "ejection occurrence" the same as "understory ejection intensity"?
114: closest cloud onset in time?
Fig 2: please use molar units for the fluxes as in Fig 1
Fig 2 caption: last sentence is vague please provide detail
122: associated instantaneous vertical wind speed
125: "anti-correlation between photosynthesis and transpiration" - the wording could be improved, these processes happen together through the stomata - you mean the fluxes go in different directions vertically (and thus their mole fractions in the atmosphere in the presence of these fluxes are anti-correlated)
129: I don't understand why you refer to Q3 as "stable background" when the entire example is stable stratification - why only Q3 and not the others?
Fig A1 in appendix: conflicting terminology here, "Q1 Respiration" (left column), but also right column says "understory ejection", but the caption says ejections are in Q1
141: seems like more than 2 events in Fig A1 rightmost column - at least 3 smaller ones there too
148: "The bulk of the flushing takes place from 7:00 to 9:00 LT, but the positive relationship between the H2O and CO2 anomalies prevents ejection events from being isolated then" - this is not what I see in that figure - there are ejection events identified prior to 09:00, and the "two phases" of ejections mentioned are not very distinct in the histogram (Fig 1D). Ejections seem to occur throughout the day and are not broken into two clear groups
151: are there observations of CBL height to compare to strengthen this argument?
Section 3.2 title: "CO2 flux partitioning" means splitting NEE into GPP and Reco - consider rewording to something like "quadrant analysis of CO2 fluxes"
156: Fig 2 has only 1 panel so no "A" needed
158: "sporadically during noon" - this is not my interpretation of Fig 2, I see sporadic events from 0900 to 1800
159: I don't understand the 3.8 to 20% numbers. The largest ejection flux in the figure has a magnitude of 0.01 mg m-2 s-1, but the net uptake of that time was ~-0.38, so the ejection flux contributes 100x 0.01/0.38 = 2.6%. Please explain your calculations in detail
183: Fig A3 does not show either Keeling or Miller-Tans plots - why even mention Keeling plots if you didn't use them?
197: "shaded understory leaves and soil evaporation are isotopically depleted compared to bulk sunlit leaves" I see 2 problems here. Most serious is that you have not measured the isotopic composition of soil *evaporation*, but bulk soil water. Second, ANOVA is needed to make the claim that understory leaves are more depleted than bulk sunlit leaves. Using the Craig-Gordon model would strengthen these claims
204: "wind ejections" is used here to refer to understory ejections, but earlier when discussing Shaw's paper you make a distinction between these - better to leave out "wind" here
210-213: this is too vague to understand
214: I would refer to this as time-lagged cross correlation (https://en.wikipedia.org/wiki/Cross-correlation)
215: I am confused about the distinction between cloud onset (start of shading) and your use of "center of clouds" here. Fig 4C makes me think that ejections happen before the cloud arrives, but you are making the opposite point I think (clouds lead to ejections). More clarity needed here
216: radiation (not radiate)
Fig 4C ejection is spelled wrong in the legend
221: the very low correlation coefficients worry me, and really weaken your argument that these are correlated - can you do more simulations to see if these values of r can be achieved randomly? maybe randomize both time series independently many times then repeat the test and compare the results
232: the synthetic simulations need more detail, I don't understand
240: more detail needed on the 1.4% - how did you calculate this?
241: moist air yes, but you do not show evidence for saturation of water vapor in air
242: "depleted" is a relative term. Understory ejections seem to have dD values of -30 to -40 permil, which is *enriched* relative to the atmospheric background (Fig 3A). The understory transpiration flux is likely close to deep soil water (-20 permil) which is also very enriched relative to background. The Craig-Gordon model would help here.
252-261: this is a wandering paragraph and rather speculative
Citation: https://doi.org/10.5194/egusphere-2025-452-RC2 - AC2: 'Reply on RC2', Robbert Moonen, 15 Jul 2025
Status: closed
-
RC1: 'Comment on egusphere-2025-452', Peter A. Taylor, 14 May 2025
While I have a basic understanding of turbulent fluxes, I am not an expert in canopy issues. Overall I see this as an interesting paper illustrating aspects of turbulent fluxes through and above the Amazon rainforest canopy.
It would be good to see some statements and ideally profile information on mixing ratios of CO2, H2O and 2H2O below, through and above the canopy in day and night time if any of these are available. The ejection concept seems to say that, above the canopy in daytime, there is generally a downward flux of CO2 but when material is ejected from below the canopy, and turbulent fluctuations CO2' and H2O' are both > 0 it is implied that the sub canopy can be a CO2 source while the canopy itself is a sink. It could be interesting to see fluxes below the canopy, as well as above, to see if there is absorption of sub-canopy CO2 as it passes through the canopy.
Nighttime seems less certain while Fig 2 shows a positive CO2 flux in early morning. The discussion in Section 3, focussed on quadrant analyses from Figures 1 and 2 is very good.
Links to cloud cover are explained and are interesting.
Detailed comments.
Abstract Was the "the depleted water vapor isotopic compositions" measured in understory air? or just leaf and soil samples?
p3 The quadrant analysis in Figure 1, and especially 1c represents the important information presented here. As I read it there should be 36,000 points in each of Figs 1a,b,c (30 min each) but only 7200 (4Hz) isotopic composition measurements. I am not sure what is meant by "a hyperbolic isolation function". Also the acronym ODR for the best fit straight lines, in the caption to Fig 1c, could be explained in section 2.1.
p3,5 The definition of an understory ejection as 0.5s with w > 0 and above a regression line, seems a little arbitrary. Were other criteria tried?
p4 Figure A3 might need more explanation, or at least a forward reference to Figure 3.
p7, 9, Fig 3, Was it a 13 day or a 14 day campaign? Could impact the number of data points averaged in Fig 3. It might also be useful to say how many days had sufficient "frequent understory ejections" in the 30 min time slots. What was the limit for data in Fig 1D - looks like about 6%.
-----------------------------------------------------
Citation: https://doi.org/10.5194/egusphere-2025-452-RC1 - AC1: 'Reply on RC1', Robbert Moonen, 10 Jul 2025
-
RC2: 'Comment on egusphere-2025-452', David Bowling, 28 May 2025
This is an excellent and novel paper focused on using high-frequency turbulence observations to evaluate the dynamic nature of rainforest canopy venting events. It is quite appropriate for the journal and generally of high quality. The datasets are quite valuable and unique and are likely to be used by others. I did not download the data but did verify that the authors have posted the data publicly already.
It appears that the authors did not extract xylem water for isotopic analysis (no data are shown). This was an unfortunate oversight for such an isotope-intensive project. For this reason, your claims about knowing the isotopic signature of understory transpiration are rather weak. I highly recommend tuning up the modern Craig-Gordon model (as described by Cernusak et al. 2016) to calculate transpiration signatures based on the data in Fig 3B.
The introduction is well written and concise, with a clear hypothesis. I have many mostly minor comments that should be addressed before publication. However, this paper definitely should be published! Nice work!
It would be great to see a version of Fig 2 for sensible and latent heat fluxes also, and maybe add typical profiles during midday of the various scalars.
Abstract: this statement is incorrect "We show that this matches the depleted water vapor isotopic compositions found in understory leaf and soil samples". Understory leaf water was about +8 permil dD, and soil water was -15 to -20 permil (Fig 3B). All of these values are more enriched than the background atmosphere and the estimated values of source composition (Fig 3A). Where is the match?
Minor comments:
3: forest atmosphere interface (not interphase)
12: weak but coherent (r=0.027) - unclear meaning upon reading abstract
14-15: how can a deeper understanding of gas exchange prevent the transition from C sink to source?
48: 20 Hz sampling rate here, but Fig 1 caption says 10 Hz
60: improved and improvement redundant
74: other usage of the hyperbolic isolation function for isotopic exchange can be found here: Bowling et al (1999) JGR 104:9121, and (2003) AFM 116:159
76: for isolation OF wind sweeps and ejections
79: 2 sigma represents standard deviation of the residuals from the ODR fit? - more mathematical detail needed here so others can reproduce this
80: ejections are Q1 not Q2 according to your notation in Fig 1 (your notation is consistent with Thomas et al. (2008) and those events were identified by them as subcanopy-venting updrafts - please be clear that the labeling of quadrants is dependent on the variables plotted - the Shaw ejections in Q2 (your Fig A5) are not the same meaning as Q2 in your Fig 1
69-72: more detail would be helpful here - it seems you are splitting the total turbulent flux based on the quadrants, and for each quadrant taking a Reynolds average of the instantaneous product of w and scalars (or w and u)? so total turbulent flux = sum of turbulent fluxes in each quadrant?
80-83: more detail needed here - show us the equations that define how you calculate the bulk and ejection fluxes
87: are those 36 consecutive points or 36 from anywhere within the 7200?
88: one-sided uncertainties is vague, and more detail would be helpful on why "unrealistic source compositions" occurred
95: incomplete sentence
108-112: this is too vague to understand - this looks like a time-lagged cross correlation between "understory ejection intensity" and "cloud intensity"??? what are "two time series" and why did you not look for the time of maximum cross-correlation? - it seems that Fig 4 D should be labeled to be consistent with these variables (is "ejection occurrence" the same as "understory ejection intensity"?
114: closest cloud onset in time?
Fig 2: please use molar units for the fluxes as in Fig 1
Fig 2 caption: last sentence is vague please provide detail
122: associated instantaneous vertical wind speed
125: "anti-correlation between photosynthesis and transpiration" - the wording could be improved, these processes happen together through the stomata - you mean the fluxes go in different directions vertically (and thus their mole fractions in the atmosphere in the presence of these fluxes are anti-correlated)
129: I don't understand why you refer to Q3 as "stable background" when the entire example is stable stratification - why only Q3 and not the others?
Fig A1 in appendix: conflicting terminology here, "Q1 Respiration" (left column), but also right column says "understory ejection", but the caption says ejections are in Q1
141: seems like more than 2 events in Fig A1 rightmost column - at least 3 smaller ones there too
148: "The bulk of the flushing takes place from 7:00 to 9:00 LT, but the positive relationship between the H2O and CO2 anomalies prevents ejection events from being isolated then" - this is not what I see in that figure - there are ejection events identified prior to 09:00, and the "two phases" of ejections mentioned are not very distinct in the histogram (Fig 1D). Ejections seem to occur throughout the day and are not broken into two clear groups
151: are there observations of CBL height to compare to strengthen this argument?
Section 3.2 title: "CO2 flux partitioning" means splitting NEE into GPP and Reco - consider rewording to something like "quadrant analysis of CO2 fluxes"
156: Fig 2 has only 1 panel so no "A" needed
158: "sporadically during noon" - this is not my interpretation of Fig 2, I see sporadic events from 0900 to 1800
159: I don't understand the 3.8 to 20% numbers. The largest ejection flux in the figure has a magnitude of 0.01 mg m-2 s-1, but the net uptake of that time was ~-0.38, so the ejection flux contributes 100x 0.01/0.38 = 2.6%. Please explain your calculations in detail
183: Fig A3 does not show either Keeling or Miller-Tans plots - why even mention Keeling plots if you didn't use them?
197: "shaded understory leaves and soil evaporation are isotopically depleted compared to bulk sunlit leaves" I see 2 problems here. Most serious is that you have not measured the isotopic composition of soil *evaporation*, but bulk soil water. Second, ANOVA is needed to make the claim that understory leaves are more depleted than bulk sunlit leaves. Using the Craig-Gordon model would strengthen these claims
204: "wind ejections" is used here to refer to understory ejections, but earlier when discussing Shaw's paper you make a distinction between these - better to leave out "wind" here
210-213: this is too vague to understand
214: I would refer to this as time-lagged cross correlation (https://en.wikipedia.org/wiki/Cross-correlation)
215: I am confused about the distinction between cloud onset (start of shading) and your use of "center of clouds" here. Fig 4C makes me think that ejections happen before the cloud arrives, but you are making the opposite point I think (clouds lead to ejections). More clarity needed here
216: radiation (not radiate)
Fig 4C ejection is spelled wrong in the legend
221: the very low correlation coefficients worry me, and really weaken your argument that these are correlated - can you do more simulations to see if these values of r can be achieved randomly? maybe randomize both time series independently many times then repeat the test and compare the results
232: the synthetic simulations need more detail, I don't understand
240: more detail needed on the 1.4% - how did you calculate this?
241: moist air yes, but you do not show evidence for saturation of water vapor in air
242: "depleted" is a relative term. Understory ejections seem to have dD values of -30 to -40 permil, which is *enriched* relative to the atmospheric background (Fig 3A). The understory transpiration flux is likely close to deep soil water (-20 permil) which is also very enriched relative to background. The Craig-Gordon model would help here.
252-261: this is a wandering paragraph and rather speculative
Citation: https://doi.org/10.5194/egusphere-2025-452-RC2 - AC2: 'Reply on RC2', Robbert Moonen, 15 Jul 2025
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 702 | 53 | 30 | 785 | 31 | 42 |
- HTML: 702
- PDF: 53
- XML: 30
- Total: 785
- BibTeX: 31
- EndNote: 42
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
While I have a basic understanding of turbulent fluxes, I am not an expert in canopy issues. Overall I see this as an interesting paper illustrating aspects of turbulent fluxes through and above the Amazon rainforest canopy.
It would be good to see some statements and ideally profile information on mixing ratios of CO2, H2O and 2H2O below, through and above the canopy in day and night time if any of these are available. The ejection concept seems to say that, above the canopy in daytime, there is generally a downward flux of CO2 but when material is ejected from below the canopy, and turbulent fluctuations CO2' and H2O' are both > 0 it is implied that the sub canopy can be a CO2 source while the canopy itself is a sink. It could be interesting to see fluxes below the canopy, as well as above, to see if there is absorption of sub-canopy CO2 as it passes through the canopy.
Nighttime seems less certain while Fig 2 shows a positive CO2 flux in early morning. The discussion in Section 3, focussed on quadrant analyses from Figures 1 and 2 is very good.
Links to cloud cover are explained and are interesting.
Detailed comments.
Abstract Was the "the depleted water vapor isotopic compositions" measured in understory air? or just leaf and soil samples?
p3 The quadrant analysis in Figure 1, and especially 1c represents the important information presented here. As I read it there should be 36,000 points in each of Figs 1a,b,c (30 min each) but only 7200 (4Hz) isotopic composition measurements. I am not sure what is meant by "a hyperbolic isolation function". Also the acronym ODR for the best fit straight lines, in the caption to Fig 1c, could be explained in section 2.1.
p3,5 The definition of an understory ejection as 0.5s with w > 0 and above a regression line, seems a little arbitrary. Were other criteria tried?
p4 Figure A3 might need more explanation, or at least a forward reference to Figure 3.
p7, 9, Fig 3, Was it a 13 day or a 14 day campaign? Could impact the number of data points averaged in Fig 3. It might also be useful to say how many days had sufficient "frequent understory ejections" in the 30 min time slots. What was the limit for data in Fig 1D - looks like about 6%.
-----------------------------------------------------