the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 May 2023
Simulating net ecosystem exchange under seasonal snow cover at an Arctic tundra site
Abstract. Estimates of winter (snow-covered non-growing season) CO2 fluxes across the Arctic region vary by a factor of three and a half, with considerable variation between measured and simulated fluxes. Measurements of snow properties, soil temperatures and net ecosystem exchange (NEE) at Trail Valley Creek, NWT, Canada, allowed evaluation of simulated winter NEE in a tundra environment with the Community Land Model (CLM5.0). Default CLM5.0 parameterisations did not adequately simulate winter NEE in this tundra environment, with near-zero NEE (< 0.01 g C m−2 d−1) simulated between November and mid-May. In contrast, measured NEE was broadly positive (indicating net CO2 release) from snow cover onset until late April. Changes to the parameterisation of snow thermal conductivity, required to correct for a cold soil temperature bias, reduced the duration for which no NEE was simulated. Parameter sensitivity analysis revealed the critical role of the minimum soil moisture threshold on decomposition (Ψmin) in regulating winter soil respiration. The default value of this parameter (Ψmin) was too high, preventing simulation of soil respiration for the vast majority of the snow-covered season. In addition, the default rate of change of soil respiration with temperature (Q10) was too low, further contributing to poor model performance during winter. As Ψmin and Q10 had opposing effects on the magnitude of simulated winter soil respiration, larger negative values of Ψmin and larger positive values of Q10 are required to simulate wintertime NEE more adequately.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(1145 KB)
-
Supplement
(347 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1145 KB) - Metadata XML
-
Supplement
(347 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-772', William Wieder, 30 Jun 2023
Dutch and co-authors look into simulations of winter time soil temperatures and heterotrophic respiration (HR) fluxes that are simulated at a EC flux site in NWT, Canada with the Community Land Model, version 5. Observations suggest positive NEE fluxes throughout much of the very cold winter, but the model simulates cold soil biases and zero HR fluxes. Building off their work looking at snow thermal properties, they explore alternative parameterizations for snow thermal conductivity, minimum soil water potential of soil organic matter decomposition, and higher temperature sensitivity (q10) of soil organic matter decomposition to capture positive NEE fluxes with the model. The paper is well written and clear.
Major concerns
I’m concerned, however, that the experimental approach seems like it is somewhat putting the cart (NEE & snow thermal properties) before the horse (simulated soil temperatures). Specifically, it seems that potential biases in soil temperature really need to be addressed before trying to adjust the sensitivity of heterotrophic respiration fluxes.
Line 187, this strikes me as a potential weakness if soil properties from the global surface datasets are not consistent with real soils at the site. It’s OK to use these data if no observational data are available, but calibrating snow thermal properties to generate warmer soils seems odd if the soil thermal and hydraulic properties are not being estimated correctly in the first place because of poor calibration to local soil properties. Specifically, I wonder if higher organic matter concentrations in surface soils would prevent the cold biases you’re seeing, especially early in the simulation?
Soils still seem to be getting pretty cold, even after modifying snow thermal conductivity parameterisations (Fig 1c). Subsequent modifications to moisture and temperature sensitivity of HR, therefore seem to be compensating for persistent soil temperature biases that ultimately make the cumulative wintertime NEE fluxes worse (shaded regions of Fig 5). Indeed, the authors settle in on really high q10 and psi min values to get any soil respiration out of the model but still have 5C temperature biases during the time of year when they’re trying to get respiration fluxes out of really cold soils.
I don’t really know what to suggest for this critique. Ideally, additional simulations could be run that either modify organic profiles on the surface dataset or explore alternative pedotransfer functions to calculate soil thermal properties. If additional simulations are not feasible, some additional discussion on addressing soil temperature biases seems warranted.
I wonder if the simulations are really spun up correctly? Specifically cumulative wintertime NEE is very high with changes to minimum soil water potential (Line 237; Fig 2). If the system is really at steady state this implies that there’s much stronger photosynthetic uptake in the growing season for these simulations. Is this true?
Minor and technical comments
Lines 53-84. This is a great literature review of some of the things that could be wrong with high latitude C fluxes that are simulated by land models, but as written it kind of gives the reader whiplash. It reads like a list of shortcomings that’s scattered around multiple processes (soil temperature, phenology and plant productivity, plus temperature and moisture sensitivity of heterotrophic respiration). Maybe breaking this up into a few paragraphs with clear topic sentences to guide the reader? I wish I had more concrete suggestions but would encourage revisions to make this text more reader friendly.
Line 64, these are both CLM references, but not “other Earth System Models”?
Line 114, what is the forcing height, ZBOT, that’s used for these single point simulations?
Line 136, was USTAR filtering done on the observations (e.g. https://ameriflux.lbl.gov/data/flux-data-products/data-qaqc/ustar-filtering-module/)?
Line 209, was the accelerated decomposition (AD) mode of the model used for this period? If so for how long? How long did you run in postAD mode? See Lawrence et al. 2019
Line 211, this is worded oddly, Lawrence et al. recommend changes in ecosystem C < 1gC/m2/y for steady state conditions.
Line 218, supplementary figures should be numbered in the order they are introduced in the text.
I find Figure 1 really hard to read. Can lines be thicker, and maybe the magnitude of the flux axes (NEE and HR) be standardized to the same scale?
I kept wondering about the evolution of snowpack that is simulated by CLM at the site. Can this be shown in Fig 1? (see also line 222 & 290). Specifically, I’d assume that snow is pretty shallow in the late-fall / early winter when surface soils are experiencing pretty cold biases (Oct-Dec, Fig 1c). Accordingly, changes to snow thermal properties has pretty minimal effects on soil temperature.
How do soil temperature profiles evolve in the simulations? Given cold surface temperature biases I’m assuming that deeper layers get quite cold too, but in reality does it take longer for these deeper horizons to freeze? Could these deeper horizons be the source of positive wintertime NEE fluxes at the site? Would this also be true in the model, or is CO2 produced deeper in CLM not allowed to move through frozen soil layers? See also Knowles et al. 2019.
To the point above, I’m surprised to see positive NEE fluxes in Feb-April when observed surface soils are so cold. I’d be surprised if microsites in surface soils at -10C could sustain such high fluxes. Can the authors speak to the physical and biological processes that are likely at play here?
LIne 245, I’m assuming this claim is in reference to 1b and the difference between when the thin red and thin blue lines hit the 0 line for HR fluxes? It’s really hard to see, but to my eye this looks like ~1 month difference. Regardless, supporting the statement with a reference to the display item seems important.
I like using display items to support claims made in the text, but none are referenced in the discussion. This is more of a stylistic comment, but it helps focus readers on highlights of the findings being presented in this study.
Line 350, this is one example of an interesting claim that could be supported by results presented in the paper?
References:
Knowles, J.F., Blanken, P.D., Lawrence, C.R. et al. Evidence for non-steady-state carbon emissions from snow-scoured alpine tundra. Nat Commun 10, 1306 (2019). https://doi.org/10.1038/s41467-019-09149-2
Citation: https://doi.org/10.5194/egusphere-2023-772-RC1 - AC1: 'Reply on RC1', Victoria Dutch, 23 Oct 2023
-
RC2: 'Comment on egusphere-2023-772', Anonymous Referee #2, 22 Sep 2023
The manuscript describes a study in which land surface model CLM5.0 was used to simulate NEE of an Arctic site, from which eddy covariance flux measurements and meteorological data were available. The default parameterisation of the model was known to produce too low wintertime NEE in Arctic environments, and the authors tested how sensitive the soil respiration simulation is to certain parameters. They found that a different parameterisation of snow thermal conductivity and lowering the minimum soil moisture required to simulate soil decomposition improved the NEE simulation most. The simulation was less sensitive to changes in Q10.
This is a clearly written manuscript that communicates nicely the results of the parameter test. Also the topic, how to improve wintertime greenhouse gas simulations in land surface models, is important. However, I think the paper content is a bit too plain and some additions are necessary. Also, I find some conclusions could be justified more. Below are my suggestions and comments.
A general comment:
You did the parameter sensitivity test at one site, Trail Valley Creek. I think the value of the study would increase a lot if you could add some Arctic site(s) and show that the new parameterisation improves wintertime NEE simulation also there, compared to the default parameters. Can you find some additional flux and met data sets from the Arctic to run your model with, e.g. via the paper you refer to: Virkkala et al. 2021?
Detailed comments:
line 22: You mention snow properties here, but in the Methods section the only snow property that you list is general snow depth (reported in King et al. 2018). Were there more? Was this information needed for the snow thermal conductivity parameterisation?
line 41: Please specify what CO2 fluxes. Non-growing season fluxes or total?
line 63-64: How well has CLM5, with the current parameterisation, simulated wintertime NEE in other than Arctic environments? E.g. boreal?
line 173: This apparently is because frozen water = ice is not part of soil moisture? Please explain it here.
line 185: Please state briefly, what the impact of the standard deviation of elevation is in this model experiment.
line 193-195: When you describe the experiment setup (2.3) you present no details about the parameterisation of snow thermal conductivity. You refer to Jordan (1991) and Sturm et al. (1997), and the reader can of course find the papers and check the details of the conductivity equation and the parameter values there, but in my opinion, it would be necessary to show them here. They are in focus, and you found that they impact the results, so you should present them also here.
line 204: The lowest Ψmin, -2000 MPa, is very low indeed. Does your model ever reach that low values, would the effect be the same if you just removed the Ψmin condition completely in the Arctic?
line 244-246: Apparently because of higher soil temperature and therefore no/less ice? Please explain it here.
line 249-250: There's a clear difference between the measured and modelled soil temperature also with the Sturm parameterisation. Please mention it here.
line 268-273 and 295-296: If I've understood correctly, you suspect that the main reason why your model overestimated NEE in early spring (and early autumn too?) was biased modelling of photosynthesis in the shoulder season? So, what is your conclusion about the greater overestimation of NEE with low Q10: is the low Q10 a worse choice although you know the overestimation also happens because of negligible/no C uptake in photosynthesis?
e.g. around line 287 (or where you think it fits): Please add some more detailed discussion explaining why and how the Sturm thermal conductivity parameter values affected the results as they did.
line 295: Why was it less rapid with larger negative Ψmin?
line 297-300 and Fig. 1: Can you partition the measured NEE and show here also the measurement-based estimate of soil respiration, not only total NEE? It would be interesting to see how well it fits with the simulation.
line 309-310 and 346: I am slightly confused by your conclusion about modelled soil moisture and the threshold. You say CLM5.0 overestimates soil moisture when soils are frozen, but on the other hand, you recommend lowering the Ψmin because with the original Ψmin (that you later describe as "overly conservative") soil decomposition ceased too easily. Please clarify this. Did the model overestimate soil moisture also in your simulation?
line 314-316: I don't understand the sentence. Can you please clarify what do you mean by "when Ψmin exceeds Ψ", how is it connected to the threshold -18°C?
line 317-320: Also here I don't understand this sentence. Can you please check if it is ok.
line 327-329: Please explain why this was. Why did you need higher Q10 with small Ψmin?
Citation: https://doi.org/10.5194/egusphere-2023-772-RC2 - AC1: 'Reply on RC1', Victoria Dutch, 23 Oct 2023
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2023-772', William Wieder, 30 Jun 2023
Dutch and co-authors look into simulations of winter time soil temperatures and heterotrophic respiration (HR) fluxes that are simulated at a EC flux site in NWT, Canada with the Community Land Model, version 5. Observations suggest positive NEE fluxes throughout much of the very cold winter, but the model simulates cold soil biases and zero HR fluxes. Building off their work looking at snow thermal properties, they explore alternative parameterizations for snow thermal conductivity, minimum soil water potential of soil organic matter decomposition, and higher temperature sensitivity (q10) of soil organic matter decomposition to capture positive NEE fluxes with the model. The paper is well written and clear.
Major concerns
I’m concerned, however, that the experimental approach seems like it is somewhat putting the cart (NEE & snow thermal properties) before the horse (simulated soil temperatures). Specifically, it seems that potential biases in soil temperature really need to be addressed before trying to adjust the sensitivity of heterotrophic respiration fluxes.
Line 187, this strikes me as a potential weakness if soil properties from the global surface datasets are not consistent with real soils at the site. It’s OK to use these data if no observational data are available, but calibrating snow thermal properties to generate warmer soils seems odd if the soil thermal and hydraulic properties are not being estimated correctly in the first place because of poor calibration to local soil properties. Specifically, I wonder if higher organic matter concentrations in surface soils would prevent the cold biases you’re seeing, especially early in the simulation?
Soils still seem to be getting pretty cold, even after modifying snow thermal conductivity parameterisations (Fig 1c). Subsequent modifications to moisture and temperature sensitivity of HR, therefore seem to be compensating for persistent soil temperature biases that ultimately make the cumulative wintertime NEE fluxes worse (shaded regions of Fig 5). Indeed, the authors settle in on really high q10 and psi min values to get any soil respiration out of the model but still have 5C temperature biases during the time of year when they’re trying to get respiration fluxes out of really cold soils.
I don’t really know what to suggest for this critique. Ideally, additional simulations could be run that either modify organic profiles on the surface dataset or explore alternative pedotransfer functions to calculate soil thermal properties. If additional simulations are not feasible, some additional discussion on addressing soil temperature biases seems warranted.
I wonder if the simulations are really spun up correctly? Specifically cumulative wintertime NEE is very high with changes to minimum soil water potential (Line 237; Fig 2). If the system is really at steady state this implies that there’s much stronger photosynthetic uptake in the growing season for these simulations. Is this true?
Minor and technical comments
Lines 53-84. This is a great literature review of some of the things that could be wrong with high latitude C fluxes that are simulated by land models, but as written it kind of gives the reader whiplash. It reads like a list of shortcomings that’s scattered around multiple processes (soil temperature, phenology and plant productivity, plus temperature and moisture sensitivity of heterotrophic respiration). Maybe breaking this up into a few paragraphs with clear topic sentences to guide the reader? I wish I had more concrete suggestions but would encourage revisions to make this text more reader friendly.
Line 64, these are both CLM references, but not “other Earth System Models”?
Line 114, what is the forcing height, ZBOT, that’s used for these single point simulations?
Line 136, was USTAR filtering done on the observations (e.g. https://ameriflux.lbl.gov/data/flux-data-products/data-qaqc/ustar-filtering-module/)?
Line 209, was the accelerated decomposition (AD) mode of the model used for this period? If so for how long? How long did you run in postAD mode? See Lawrence et al. 2019
Line 211, this is worded oddly, Lawrence et al. recommend changes in ecosystem C < 1gC/m2/y for steady state conditions.
Line 218, supplementary figures should be numbered in the order they are introduced in the text.
I find Figure 1 really hard to read. Can lines be thicker, and maybe the magnitude of the flux axes (NEE and HR) be standardized to the same scale?
I kept wondering about the evolution of snowpack that is simulated by CLM at the site. Can this be shown in Fig 1? (see also line 222 & 290). Specifically, I’d assume that snow is pretty shallow in the late-fall / early winter when surface soils are experiencing pretty cold biases (Oct-Dec, Fig 1c). Accordingly, changes to snow thermal properties has pretty minimal effects on soil temperature.
How do soil temperature profiles evolve in the simulations? Given cold surface temperature biases I’m assuming that deeper layers get quite cold too, but in reality does it take longer for these deeper horizons to freeze? Could these deeper horizons be the source of positive wintertime NEE fluxes at the site? Would this also be true in the model, or is CO2 produced deeper in CLM not allowed to move through frozen soil layers? See also Knowles et al. 2019.
To the point above, I’m surprised to see positive NEE fluxes in Feb-April when observed surface soils are so cold. I’d be surprised if microsites in surface soils at -10C could sustain such high fluxes. Can the authors speak to the physical and biological processes that are likely at play here?
LIne 245, I’m assuming this claim is in reference to 1b and the difference between when the thin red and thin blue lines hit the 0 line for HR fluxes? It’s really hard to see, but to my eye this looks like ~1 month difference. Regardless, supporting the statement with a reference to the display item seems important.
I like using display items to support claims made in the text, but none are referenced in the discussion. This is more of a stylistic comment, but it helps focus readers on highlights of the findings being presented in this study.
Line 350, this is one example of an interesting claim that could be supported by results presented in the paper?
References:
Knowles, J.F., Blanken, P.D., Lawrence, C.R. et al. Evidence for non-steady-state carbon emissions from snow-scoured alpine tundra. Nat Commun 10, 1306 (2019). https://doi.org/10.1038/s41467-019-09149-2
Citation: https://doi.org/10.5194/egusphere-2023-772-RC1 - AC1: 'Reply on RC1', Victoria Dutch, 23 Oct 2023
-
RC2: 'Comment on egusphere-2023-772', Anonymous Referee #2, 22 Sep 2023
The manuscript describes a study in which land surface model CLM5.0 was used to simulate NEE of an Arctic site, from which eddy covariance flux measurements and meteorological data were available. The default parameterisation of the model was known to produce too low wintertime NEE in Arctic environments, and the authors tested how sensitive the soil respiration simulation is to certain parameters. They found that a different parameterisation of snow thermal conductivity and lowering the minimum soil moisture required to simulate soil decomposition improved the NEE simulation most. The simulation was less sensitive to changes in Q10.
This is a clearly written manuscript that communicates nicely the results of the parameter test. Also the topic, how to improve wintertime greenhouse gas simulations in land surface models, is important. However, I think the paper content is a bit too plain and some additions are necessary. Also, I find some conclusions could be justified more. Below are my suggestions and comments.
A general comment:
You did the parameter sensitivity test at one site, Trail Valley Creek. I think the value of the study would increase a lot if you could add some Arctic site(s) and show that the new parameterisation improves wintertime NEE simulation also there, compared to the default parameters. Can you find some additional flux and met data sets from the Arctic to run your model with, e.g. via the paper you refer to: Virkkala et al. 2021?
Detailed comments:
line 22: You mention snow properties here, but in the Methods section the only snow property that you list is general snow depth (reported in King et al. 2018). Were there more? Was this information needed for the snow thermal conductivity parameterisation?
line 41: Please specify what CO2 fluxes. Non-growing season fluxes or total?
line 63-64: How well has CLM5, with the current parameterisation, simulated wintertime NEE in other than Arctic environments? E.g. boreal?
line 173: This apparently is because frozen water = ice is not part of soil moisture? Please explain it here.
line 185: Please state briefly, what the impact of the standard deviation of elevation is in this model experiment.
line 193-195: When you describe the experiment setup (2.3) you present no details about the parameterisation of snow thermal conductivity. You refer to Jordan (1991) and Sturm et al. (1997), and the reader can of course find the papers and check the details of the conductivity equation and the parameter values there, but in my opinion, it would be necessary to show them here. They are in focus, and you found that they impact the results, so you should present them also here.
line 204: The lowest Ψmin, -2000 MPa, is very low indeed. Does your model ever reach that low values, would the effect be the same if you just removed the Ψmin condition completely in the Arctic?
line 244-246: Apparently because of higher soil temperature and therefore no/less ice? Please explain it here.
line 249-250: There's a clear difference between the measured and modelled soil temperature also with the Sturm parameterisation. Please mention it here.
line 268-273 and 295-296: If I've understood correctly, you suspect that the main reason why your model overestimated NEE in early spring (and early autumn too?) was biased modelling of photosynthesis in the shoulder season? So, what is your conclusion about the greater overestimation of NEE with low Q10: is the low Q10 a worse choice although you know the overestimation also happens because of negligible/no C uptake in photosynthesis?
e.g. around line 287 (or where you think it fits): Please add some more detailed discussion explaining why and how the Sturm thermal conductivity parameter values affected the results as they did.
line 295: Why was it less rapid with larger negative Ψmin?
line 297-300 and Fig. 1: Can you partition the measured NEE and show here also the measurement-based estimate of soil respiration, not only total NEE? It would be interesting to see how well it fits with the simulation.
line 309-310 and 346: I am slightly confused by your conclusion about modelled soil moisture and the threshold. You say CLM5.0 overestimates soil moisture when soils are frozen, but on the other hand, you recommend lowering the Ψmin because with the original Ψmin (that you later describe as "overly conservative") soil decomposition ceased too easily. Please clarify this. Did the model overestimate soil moisture also in your simulation?
line 314-316: I don't understand the sentence. Can you please clarify what do you mean by "when Ψmin exceeds Ψ", how is it connected to the threshold -18°C?
line 317-320: Also here I don't understand this sentence. Can you please check if it is ok.
line 327-329: Please explain why this was. Why did you need higher Q10 with small Ψmin?
Citation: https://doi.org/10.5194/egusphere-2023-772-RC2 - AC1: 'Reply on RC1', Victoria Dutch, 23 Oct 2023
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 426 | 174 | 25 | 625 | 51 | 15 | 18 |
- HTML: 426
- PDF: 174
- XML: 25
- Total: 625
- Supplement: 51
- BibTeX: 15
- EndNote: 18
Viewed (geographical distribution)
| Country | # | Views | % |
|---|---|---|---|
| United States of America | 1 | 180 | 29 |
| United Kingdom | 2 | 143 | 23 |
| Germany | 3 | 55 | 8 |
| Canada | 4 | 42 | 6 |
| China | 5 | 29 | 4 |
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
- 180
Victoria R. Dutch
Nick Rutter
Leanne Wake
Oliver Sonnentag
Gabriel Hould Gosselin
Melody Sandells
Chris Derksen
Branden Walker
Gesa Meyer
Richard Essery
Richard Kelly
Phillip Marsh
Julia Boike
Matteo Detto
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(1145 KB) - Metadata XML
-
Supplement
(347 KB) - BibTeX
- EndNote
- Final revised paper