Process-based analysis of terrestrial carbon flux predictability
- 1Max Planck Institute for Meteorology, Hamburg, Germany
- 2International Max Planck Research School on Earth System Modelling, Hamburg, Germany
- 3College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, UK
- 4Center for Earth System Research and Sustainability, University of Hamburg, Germany
- 1Max Planck Institute for Meteorology, Hamburg, Germany
- 2International Max Planck Research School on Earth System Modelling, Hamburg, Germany
- 3College of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, UK
- 4Center for Earth System Research and Sustainability, University of Hamburg, Germany
Abstract. Despite efforts to decrease the discrepancy between simulated and observed terrestrial carbon fluxes, the uncertainty in trends and patterns of the land carbon fluxes remains high. This difficulty raises the question to what extent the terrestrial carbon cycle is predictable, and which processes explain the predictability. Here, the perfect model approach is used to assess the potential predictability of net primary production (NPPpred) and heterotrophic respiration (Rhpred) by using ensemble simulations conducted with the Max-Planck-Institute Earth System Model. In order to asses the role of local carbon flux predictability (CFpred) on the predictability of the global carbon cycle, we suggest a new predictability metric weighted by the amplitude of the flux anomalies. Regression analysis is used to determine the contribution of the predictability of different environmental drivers to NPPpred and Rhpred (soil moisture, air temperature and radiation for NPP and soil organic carbon, air temperature and precipitation for Rh). NPPpred is driven to 62 and 30 % by the predictability of soil moisture and temperature, respectively. Rhpred is driven to 52 and 27 % by the predictability of soil organic carbon temperature, respectively. The decomposition of predictability shows that the relatively high Rhpred compared to NPPpred is due to the generally high predictability of soil organic carbon. The seasonality in NPPpred and Rhpred patterns can be explained by the change in limiting factors over the wet and dry months. Consequently, CFpred is controlled by the predictability of the currently limiting environmental factor. Differences in CFpred between ensemble simulations can be attributed to the occurrence of wet and dry years, which influences the predictability of soil moisture and temperature. This variability of predictability is caused by the state dependency of ecosystem processes. Our results reveal the crucial regions and ecosystem processes to be considered when initializing a carbon prediction system.
-
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
(8749 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Journal article(s) based on this preprint
István Dunkl et al.
Interactive discussion
Status: closed
-
RC1: 'Comment on esd-2021-38', Anonymous Referee #1, 11 Aug 2021
Review of: Process-based analysis of terrestrial carbon flux predictability
Dunkl et al.
Overview:
The paper uses a perfect model approach and three statistical metrics to investigate the predictability of carbon fluxes at an annual time scale. The study finds that the spatial-temporal pattern of carbon flux predictability is governed by complex contributions from ENSO, seasonal limiting factors and non-linear ecosystem responses. The paper is generally well written, scientifically sound, and an important contribution to understanding carbon fluxes. I recommend that the paper be accepted with minor revisions.General Comments:
1) The 'perfect model framework' is mentioned in the introduction long before it is explained in the methods section. Adding a paragraph to the introduction to explain what the method is, where it has been used before, and its limitations, would help frame the paper better.
2) In the discussion section a paragraph could be added to explain the practical implications of the results. This is briefly mentioned in the last line of the conclusions but it could be fleshed-out better. As is, the paper does not do a good job of explaining why readers should care about the results.
Specific Comments:
Line 21: Be clearer here about whether you mean the seasonal cycle or annual variation in the 1st derivative of CO2 concentration.
Line 29: Unclear what "of emission reduction detection in the face of internal variability" means.
Line 39: Change 'here' to 'therein'
Line 68: An abbreviation for standard deviation seems unnecessary. Also was the abbreviation ever introduced?
Line 97: 'verification' is the wrong word to use here.
Figure 3: Rephrase caption to eliminate 'significant'. 'values above the 95\% confidence ...' is good enough to convey the meaning with stepping on the land-mine of whether or not statistical significance is a metric that should exist.
Figure 7: Explain the yellow triangle in the figure caption.
-
AC1: 'Reply on RC1', István Dunkl, 09 Sep 2021
1. The 'perfect model framework' is mentioned in the introduction long before it is explained in the methods section. Adding a paragraph to the introduction to explain what the method is, where it has been used before, and its limitations, would help frame the paper better.
We acknowledge the lack of description of this key method at an early point in the manuscript and are adding some context on the perfect model framework to the introduction.
2 In the discussion section a paragraph could be added to explain the practical implications of the results. This is briefly mentioned in the last line of the conclusions but it could be fleshed-out better. As is, the paper does not do a good job of explaining why readers should care about the results.
The main focus of the discussion is on the nature of the findings and some outlook is indeed needed. We are adding this to the discussion.
Line 21: Be clearer here about whether you mean the seasonal cycle or annual variation in the 1st derivative of CO2 concentration.
To our understanding interannual is the term used to describe the differences between years, while intra-annual would be used for variations within years (season to season).
Line 29: Unclear what "of emission reduction detection in the face of internal variability" means.
The phrase refers to the problem of detecting the effect of of carbon emission reduction policies by measuring the atmospheric CO2 concentration. This is the case because small changes in emissions are overshadowed by the large natural variability of atmospheric CO2 concentrations.
Line 39: Change 'here' to 'therein'
Text will be changed accordingly.
Line 68: An abbreviation for standard deviation seems unnecessary. Also was the abbreviation ever introduced?
The abbreviation is introduced in line 22. We find it to be a useful and common abbreviation.
Line 97: 'verification' is the wrong word to use here.
We thank the reviewer to point out the mistake and we change the phrase to ‘validation’.
Figure 3: Rephrase caption to eliminate 'significant'. 'values above the 95\% confidence...' is good enough to convey the meaning with stepping on the land-mine of whether or not statistical significance is a metric that should exist.
Text will be changed accordingly.
Figure 7: Explain the yellow triangle in the figure caption.
The legends to this figure were incomplete and will be extended.
-
AC1: 'Reply on RC1', István Dunkl, 09 Sep 2021
-
RC2: 'Comment on esd-2021-38', Anonymous Referee #2, 09 Sep 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2021-38/esd-2021-38-RC2-supplement.pdf
-
AC2: 'Reply on RC2', István Dunkl, 09 Sep 2021
1. The selected variables for NPP and Rh have good references and reasons, and it may need some explanations/discussions why not precipitation and CO2 [1] for NPP, and why not soil moisture, soil clay content [2, 3] (important for soil respiration) for Rh, and the different/related effects in precipitation and soil moisture for NPP and Rh (e.g. the time lag effect of soil moisture with precipitation).
Both NPP and Rh are processes which are highly dependent on water availability. While the vegetation responsible for NPP can access multiple soil layers, most of Rh is taking place in the litter and the topmost soil layer. The moisture dynamics in these layers are more closely related to monthly precipitation then the moisture in the whole soil column (we will add a clarifying sentence on this in the manuscript). Additionally, the soil respiration sub-model is also using precipitation to calculate the rate of Rh.
CO2 fertilization plays a large role in the prediction of the carbon flux trend. However we assumed that the interannual variability of near surface atmospheric CO2, which has an effect on seasonal to decadal predictability, is low.
Clay content is not considered by the Rh model used in MPI-ESM [1]. If the module would include clay content in the calculation of soil respiration, its effect could be measurable through the predictability of SOC, because clay content directly effects SOC dynamics.
2. Are there any conditions for the results of 62% for soil moisture for NPPpred and 52% for SOC for Rhpred (add words that this is for global mean, and discuss with key regions such as Amazon)? And it needs to be more specific for “reveal the crucial regions and ecosystem processes to be considered when initializing a carbon prediction system”.
These numbers have been used without sufficient context in the abstract. The needed context is added to the manuscript.
We think that elaborating on the regional patterns and predictability hotspots would be beyond the scope of the abstract. These details are mentioned in the chapters 3.1 and 3.2.
3. The scale mismatch problem between site observed data and model simulated results makes the comparison of NPP and Rh very difficult, and thus result the difficulties in reducing uncertainty in simulated terrestrial carbon fluxes. And this raises some questions on true meaning of calibrating models with site specific observations with several sets of parameters and their spatial representatives (line 30). Such mismatch may deserve discussions. And I cannot find the o (validation anomalies) descriptions for global gridded NPP and Rh. And some discussions of uncertainties in model structures such as the models involved in TRENDY may be needed.
The reviewer is mentioning the difficulties arising from model parameterization based on site observations. We agree on the mentioned points, however, the validation of the carbon flux patterns is not within the interest of this study. The referred section in the text is included as an example for the efforts being made bring modelled processes closer to observations.
1. Add “and” in line 10 between “soil organic carbon” and “temperature”.
Added.
2. Extend implications of this study, for example, can the results here help to constrain the uncertainty in land sink projections?
We acknowledge the limitations of the discussion in this regard and will add further outlook.
3. Can add this ref Zeng et al., (2014) [4] in refine model structure (line 31-32);
The publication is a good example of refined models to improve the simulation of the terrestrial carbon cycle and we can add it to the list.
4. Explain somewhat of “the perfect model framework” in line 36, and why is it called “perfect”?;
Indeed, the description of the term should belong in the introduction and not the methods. We will change the text accordingly.
5. Why Fig.2, 5 and 6 only showed -30~30 instead of -90~90?;
These plots show a subset of the tropics and sub-tropics where the highest potential for (long-term) predictability is located. We will mention the reason for the subsetting in the main text.
6. What are “other factors” in Line 169; And why the Congo basin is not strongly affected by ENSO?
We recognize that “other factors” sounds missleading and will rewrite this sentence.
7. Fig.7 needs legend for black rectangle and yellow triangle and relevance with the following figures and analyses;
A proper description will be added to the figure caption.
8. The long term effects of the initial soil moisture would become very weak for Fig.7? And blue color means lower NPP predictability in wet years in Fig.7 ?
The long lasting effects of these anomalies are because a) they include only the 20% tail end of all extreme years (dry and wet), meaning that these anomalies are of such a strong magnitude that their effect will remain for several months, and b) many of the extreme years are strong El Niño or La Niña years with sustained precipitation anomalies that will elongate the soil moisture anomalies. The description of blue colors will be added to the figure captions.
9. Are there mechanisms in switch of deepSOIL and midSOIL for La Nina in Fig.8 from March to June?
As the dry season begins with the boreal summer, the topsoil dries out and vegetation growth is increasingly driven by moisture from deeper soil layers.
10. Line 296, the driving factors can be different across key regions (such as discussions in Lines 169), can add some specific summary on key regions.
We will elaborate on the discussion of the regional patterns.
11. Line 413, delete space of “CO 2”;
Text will be modified accordingly.
12. Lines 375-426, need to maintain reference formats such as to capitalize journal names (e.g. Functional plant biology; Global change biology; Global biogeochemical cycles).
A uniform format will be applied.
[1] Tuomi, Mikko, et al. "Leaf litter decomposition—Estimates of global variability based on Yasso07 model." Ecological Modelling 220.23 (2009): 3362-3371.
-
AC2: 'Reply on RC2', István Dunkl, 09 Sep 2021
Peer review completion
Interactive discussion
Status: closed
-
RC1: 'Comment on esd-2021-38', Anonymous Referee #1, 11 Aug 2021
Review of: Process-based analysis of terrestrial carbon flux predictability
Dunkl et al.
Overview:
The paper uses a perfect model approach and three statistical metrics to investigate the predictability of carbon fluxes at an annual time scale. The study finds that the spatial-temporal pattern of carbon flux predictability is governed by complex contributions from ENSO, seasonal limiting factors and non-linear ecosystem responses. The paper is generally well written, scientifically sound, and an important contribution to understanding carbon fluxes. I recommend that the paper be accepted with minor revisions.General Comments:
1) The 'perfect model framework' is mentioned in the introduction long before it is explained in the methods section. Adding a paragraph to the introduction to explain what the method is, where it has been used before, and its limitations, would help frame the paper better.
2) In the discussion section a paragraph could be added to explain the practical implications of the results. This is briefly mentioned in the last line of the conclusions but it could be fleshed-out better. As is, the paper does not do a good job of explaining why readers should care about the results.
Specific Comments:
Line 21: Be clearer here about whether you mean the seasonal cycle or annual variation in the 1st derivative of CO2 concentration.
Line 29: Unclear what "of emission reduction detection in the face of internal variability" means.
Line 39: Change 'here' to 'therein'
Line 68: An abbreviation for standard deviation seems unnecessary. Also was the abbreviation ever introduced?
Line 97: 'verification' is the wrong word to use here.
Figure 3: Rephrase caption to eliminate 'significant'. 'values above the 95\% confidence ...' is good enough to convey the meaning with stepping on the land-mine of whether or not statistical significance is a metric that should exist.
Figure 7: Explain the yellow triangle in the figure caption.
-
AC1: 'Reply on RC1', István Dunkl, 09 Sep 2021
1. The 'perfect model framework' is mentioned in the introduction long before it is explained in the methods section. Adding a paragraph to the introduction to explain what the method is, where it has been used before, and its limitations, would help frame the paper better.
We acknowledge the lack of description of this key method at an early point in the manuscript and are adding some context on the perfect model framework to the introduction.
2 In the discussion section a paragraph could be added to explain the practical implications of the results. This is briefly mentioned in the last line of the conclusions but it could be fleshed-out better. As is, the paper does not do a good job of explaining why readers should care about the results.
The main focus of the discussion is on the nature of the findings and some outlook is indeed needed. We are adding this to the discussion.
Line 21: Be clearer here about whether you mean the seasonal cycle or annual variation in the 1st derivative of CO2 concentration.
To our understanding interannual is the term used to describe the differences between years, while intra-annual would be used for variations within years (season to season).
Line 29: Unclear what "of emission reduction detection in the face of internal variability" means.
The phrase refers to the problem of detecting the effect of of carbon emission reduction policies by measuring the atmospheric CO2 concentration. This is the case because small changes in emissions are overshadowed by the large natural variability of atmospheric CO2 concentrations.
Line 39: Change 'here' to 'therein'
Text will be changed accordingly.
Line 68: An abbreviation for standard deviation seems unnecessary. Also was the abbreviation ever introduced?
The abbreviation is introduced in line 22. We find it to be a useful and common abbreviation.
Line 97: 'verification' is the wrong word to use here.
We thank the reviewer to point out the mistake and we change the phrase to ‘validation’.
Figure 3: Rephrase caption to eliminate 'significant'. 'values above the 95\% confidence...' is good enough to convey the meaning with stepping on the land-mine of whether or not statistical significance is a metric that should exist.
Text will be changed accordingly.
Figure 7: Explain the yellow triangle in the figure caption.
The legends to this figure were incomplete and will be extended.
-
AC1: 'Reply on RC1', István Dunkl, 09 Sep 2021
-
RC2: 'Comment on esd-2021-38', Anonymous Referee #2, 09 Sep 2021
The comment was uploaded in the form of a supplement: https://esd.copernicus.org/preprints/esd-2021-38/esd-2021-38-RC2-supplement.pdf
-
AC2: 'Reply on RC2', István Dunkl, 09 Sep 2021
1. The selected variables for NPP and Rh have good references and reasons, and it may need some explanations/discussions why not precipitation and CO2 [1] for NPP, and why not soil moisture, soil clay content [2, 3] (important for soil respiration) for Rh, and the different/related effects in precipitation and soil moisture for NPP and Rh (e.g. the time lag effect of soil moisture with precipitation).
Both NPP and Rh are processes which are highly dependent on water availability. While the vegetation responsible for NPP can access multiple soil layers, most of Rh is taking place in the litter and the topmost soil layer. The moisture dynamics in these layers are more closely related to monthly precipitation then the moisture in the whole soil column (we will add a clarifying sentence on this in the manuscript). Additionally, the soil respiration sub-model is also using precipitation to calculate the rate of Rh.
CO2 fertilization plays a large role in the prediction of the carbon flux trend. However we assumed that the interannual variability of near surface atmospheric CO2, which has an effect on seasonal to decadal predictability, is low.
Clay content is not considered by the Rh model used in MPI-ESM [1]. If the module would include clay content in the calculation of soil respiration, its effect could be measurable through the predictability of SOC, because clay content directly effects SOC dynamics.
2. Are there any conditions for the results of 62% for soil moisture for NPPpred and 52% for SOC for Rhpred (add words that this is for global mean, and discuss with key regions such as Amazon)? And it needs to be more specific for “reveal the crucial regions and ecosystem processes to be considered when initializing a carbon prediction system”.
These numbers have been used without sufficient context in the abstract. The needed context is added to the manuscript.
We think that elaborating on the regional patterns and predictability hotspots would be beyond the scope of the abstract. These details are mentioned in the chapters 3.1 and 3.2.
3. The scale mismatch problem between site observed data and model simulated results makes the comparison of NPP and Rh very difficult, and thus result the difficulties in reducing uncertainty in simulated terrestrial carbon fluxes. And this raises some questions on true meaning of calibrating models with site specific observations with several sets of parameters and their spatial representatives (line 30). Such mismatch may deserve discussions. And I cannot find the o (validation anomalies) descriptions for global gridded NPP and Rh. And some discussions of uncertainties in model structures such as the models involved in TRENDY may be needed.
The reviewer is mentioning the difficulties arising from model parameterization based on site observations. We agree on the mentioned points, however, the validation of the carbon flux patterns is not within the interest of this study. The referred section in the text is included as an example for the efforts being made bring modelled processes closer to observations.
1. Add “and” in line 10 between “soil organic carbon” and “temperature”.
Added.
2. Extend implications of this study, for example, can the results here help to constrain the uncertainty in land sink projections?
We acknowledge the limitations of the discussion in this regard and will add further outlook.
3. Can add this ref Zeng et al., (2014) [4] in refine model structure (line 31-32);
The publication is a good example of refined models to improve the simulation of the terrestrial carbon cycle and we can add it to the list.
4. Explain somewhat of “the perfect model framework” in line 36, and why is it called “perfect”?;
Indeed, the description of the term should belong in the introduction and not the methods. We will change the text accordingly.
5. Why Fig.2, 5 and 6 only showed -30~30 instead of -90~90?;
These plots show a subset of the tropics and sub-tropics where the highest potential for (long-term) predictability is located. We will mention the reason for the subsetting in the main text.
6. What are “other factors” in Line 169; And why the Congo basin is not strongly affected by ENSO?
We recognize that “other factors” sounds missleading and will rewrite this sentence.
7. Fig.7 needs legend for black rectangle and yellow triangle and relevance with the following figures and analyses;
A proper description will be added to the figure caption.
8. The long term effects of the initial soil moisture would become very weak for Fig.7? And blue color means lower NPP predictability in wet years in Fig.7 ?
The long lasting effects of these anomalies are because a) they include only the 20% tail end of all extreme years (dry and wet), meaning that these anomalies are of such a strong magnitude that their effect will remain for several months, and b) many of the extreme years are strong El Niño or La Niña years with sustained precipitation anomalies that will elongate the soil moisture anomalies. The description of blue colors will be added to the figure captions.
9. Are there mechanisms in switch of deepSOIL and midSOIL for La Nina in Fig.8 from March to June?
As the dry season begins with the boreal summer, the topsoil dries out and vegetation growth is increasingly driven by moisture from deeper soil layers.
10. Line 296, the driving factors can be different across key regions (such as discussions in Lines 169), can add some specific summary on key regions.
We will elaborate on the discussion of the regional patterns.
11. Line 413, delete space of “CO 2”;
Text will be modified accordingly.
12. Lines 375-426, need to maintain reference formats such as to capitalize journal names (e.g. Functional plant biology; Global change biology; Global biogeochemical cycles).
A uniform format will be applied.
[1] Tuomi, Mikko, et al. "Leaf litter decomposition—Estimates of global variability based on Yasso07 model." Ecological Modelling 220.23 (2009): 3362-3371.
-
AC2: 'Reply on RC2', István Dunkl, 09 Sep 2021
Peer review completion
Journal article(s) based on this preprint
István Dunkl et al.
István Dunkl et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
516 | 196 | 15 | 727 | 7 | 12 |
- HTML: 516
- PDF: 196
- XML: 15
- Total: 727
- BibTeX: 7
- EndNote: 12
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1
Cited
1 citations as recorded by crossref.
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(8749 KB) - Metadata XML