the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 21 Jan 2025
Mapping global distributions, environmental controls, and uncertainties of apparent top- and subsoil organic carbon turnover times
Abstract. The turnover time (τ) of global soil organic carbon is central to the functioning of terrestrial ecosystems. Yet our spatially-explicit understanding of depth-dependent variations and environmental controls of τ at a global scale remain incomplete. In this study, we combine multiple state-of-the-art observation-based datasets, including over ninety thousand geo-referenced soil profiles, the latest root observations distributed globally, and large amounts of satellite-derived environmental variables, to generate global maps of apparent τ in topsoil (0–0.3 m) and subsoil (0.3–1 m) layers with a spatial resolution of 30 arcsec (~1 km at the Equator). We show that subsoil τ (385203485 years [mean with a variation range from 2.5th to 97.5th percentile]) is over eight times longer than topsoil τ (1511137 years). The cross-validation shows that the fitted machine learning models effectively captured the variabilities in τ, with R2 values of 0.87 and 0.70 for topsoil and subsoil τ mapping, respectively. The prediction uncertainties of the τ maps were quantified for better user applications. The environmental controls on top- and subsoil τ were investigated at global, biome, and local scales. Our analyses illustrate that how temperature, water availability, physio-chemical properties and depth exert jointly impacts on τ. The data-driven approaches allow us to identify their interactions, thereby enriching our comprehension of mechanisms driving nonlinear τ–environment relationships from global to local scales. The distributions of dominating factors of τ at local scales were mapped for identifying context-dependent controls on τ across different regions. We further reveal that the current Earth system models may underestimate τ by comparing model-derived maps with our observation-derived τ maps. The resulting maps with new insights demonstrated in this study facilitate the future modelling efforts of carbon cycle–climate feedbacks and supporting effective carbon management. The dataset is archived and freely available at https://doi.org/10.5281/zenodo.14560239 (Zhang, 2025).
- Preprint
(7609 KB) - Metadata XML
-
Supplement
(2561 KB) - BibTeX
- EndNote
Status: closed
-
RC1: 'Comment on essd-2024-618', Anonymous Referee #1, 03 Feb 2025
This study creates global maps of soil organic carbon turnover times (τ) at both topsoil and subsoil depths with observation-based datasets and machine learning methods. It addresses an important knowledge gap regarding the global maps related to the depth-dependent variations and environmental controls of soil carbon turnover, which is essential for understanding terrestrial carbon storage and dynamics, especially in the context of climate change. The methods used are well-suited to the scale and complexity of the question being addressed. The use of over ninety thousand geo-referenced soil profiles, satellite-derived environmental variables, and machine learning models, provides a robust framework for mapping and analyzing τ across a wide range of environmental conditions and different biomes. The authors’ approach to quantify the uncertainty of their maps also adds considerable value, making their results more applicable to future carbon cycle modelling and land management efforts. In conclusion, I recommend the acceptance of this manuscript after revisions following several suggestions I listed below.
(1) Organic carbon turnover times can be calculated from influencing factors, such as carbon allocation belowground, SOC stocks. Why not first generate the spatial distribution data of those factors based on sampling data with machine learning method, then calculate organic carbon turnover times with the physical equations.
(2) From the physical equations 1 to 8, the importance of different factors could be derived. Do the importance results of RF coincide with those equations?
(3) While the methods section is thorough, some readers may benefit from further clarification on the choice of specific methods, particularly regarding the machine learning model calibration. A more explicit discussion of how the model’s hyper-parameters were tuned would help readers better appreciate the methodology’s rigor.
(4) The cross-validation procedure is well-executed, but it would be helpful to provide more specific details on how the biome-specific samples were handled during the 10-fold validation process. Ensuring that the sample division properly accounts for the geographical and ecological variation across biomes could further strengthen the model’s credibility.
(5) The paper requires careful attention to language for clarity and readability. I would recommend reviewing some descriptions for possible simplifications in sentence structure and corrections in grammar. Some examples are listed below.
- 18: “… is central to …” -> “… plays a crucial role in …”.
- 19: “remain” -> “remains”.
- 92: it is better to change the word “minimize” to “reduce”.
- 103: “layers observations” -> “layer observations”.
- 465: “insights of this study -> “insights from this study”.
Citation: https://doi.org/10.5194/essd-2024-618-RC1 - AC1: 'Reply on RC1', Lei Zhang, 14 Mar 2025
-
RC2: 'Comment on essd-2024-618', Anonymous Referee #2, 22 Feb 2025
General comments:
This manuscript calculated the apparent turnover time of top and subsoil SOC on a global scale. The major outcome from this work is very useful and timely needed for soil biogeochemistry and carbon cycle modeling communities. The comprehensive data inputs authors used, random forest based geospatial predictive mapping and in-depth uncertainty analysis guaranteed the quality of the produced τ map. Overall, this is a solid and interesting work and I would like to recommend it for acceptance after some technical revisions.
Specific comments:
- The description of quantile regression forest is not very straightforward. Can authors double check and revise this part?
- There lacks certain discussion of the topographical effect on τ. Since the topography largely impacts τ in tundra covered regions (Fig. S23, S24), can authors discuss this finding?
Technical corrections:
I suggest authors double check and add units to all variables in your equations.
Line 89: "collected form" shall be "collected from"
Line 99 - 100: In equation 1, SOC_Du-Dl is not defined.
Line 100: "Were" shall be "Where"
Line 103: "fit layers observations at different depth intervals" can be simplified "fit observations along depth"?
Line 105: "the SOCS at two layers for" shall be more specific "the SOCS at top- and subsoil layers for"
Line 112: "The flux of carbon at a certain soil layer needs to be obtained through" can be revised "Carbon influx at each soil layer comes from"
Line 114: "The annual NPP (kg C m-2 yr-1) produced by the moderate-resolution imaging spectroradiometer (MODIS)" means the MOD17 products? Please add the specific version of the MOD17 product and the url where the author downloaded this data.
Line 126: "total amount of roots" shall be "total root biomass"?
Line 132: " the root distribution" shall be " the root biomass distribution"?
Line 144: " for each soil sample site, root profile observations within the same terrestrial ecoregions (Dinerstein et al., 2017) and the same soil type (FAO–Unesco, 1990) as that of the soil sample were selected. The corresponding mean of those selected root observations for each soil sample location were finally collected (Figs. S8 and S9)" can be simplified "we apply the arithmetic mean of fr from root profile observations within the same terrestrial ecoregions (Dinerstein et al., 2017) and soil type (FAO–Unesco, 1990) as soil sample."
Line 189: "The partial correlation of each influencing factor was calculated while controlling other factors" at which level? Mean or median?
Line 201: "and this division was performed on each biome data to ensure that the ten split sets can keep a balance among biomes". If my understanding is correct, do you mean "and samples of each biome in each subset has the same proportion as the whole dataset"?
Line 209: "by replacing observations by indicator transforms". Not sure I understand "indicator transforms". Do you mean replacing the actual observation values by a tag of category? If so I'm not sure why this is needed. Can you provide a simple example?
Line 213: "has been also" shall be "has also been"
Line 215: If my understanding is correct, it's nice to consider error propagation and produce a dataset considering this uncertainty information to train quantile regression forests. But in your writing this information is not explicitly conveyed so I feel a bit confused when reading this part. It will be better if authors can explicitly tell readers this information above this paragraph.
Line 250: Unfinished sentence "The shaded grey area represents the"
Figure 6: is interesting. I guess the y axis title "MAT/MAP effect" means the partial regression coefficient between MAT/MAP and τ? Please add a description in figure caption.
Figure 6: Another question. I know to completely decorrelate climate and edaphic variables is extremely tough, but I would suggest authors provide a correlation matrix to visualize and identify potential issues with "multicollinearity" between 2 climate and 4 edaphic variables.
Figure 6d shows different responses for topsoil and subsoil. As high CEC enhances adsorption, more nutrients would be available for micro-organism and more mineral-organic compounds are formed, which tends to increase sensitivity of SOCS to temperature. But meanwhile, plants are more productive and may have higher plant carbon stock, thus might increase sensitivity of NPP to temperature. The results might be the combination of both effects.
Figure 7c: What is the unit of variable importance (%)?
Line 405: "biogeochemical simulations by ESMs, and will be useful to improve" can be simplified to "ESM simulations and improve"
Line 407: shall mention CMIP6 outputs from which experiment?
From Fig. S17, it seems like most of the wetland samples are collected from the tropics. Would you justify the performance of wetland τ the arctic region?
Citation: https://doi.org/10.5194/essd-2024-618-RC2 - AC2: 'Reply on RC2', Lei Zhang, 14 Mar 2025
-
RC3: 'Comment on essd-2024-618', Anonymous Referee #3, 05 Mar 2025
The manuscript titled “Mapping global distributions, environmental controls, and uncertainties of apparent top- and subsoil organic carbon turnover times” addresses an important knowledge gap and is therefore likely to be of broad interest. The primary aim of the study is to map carbon turnover times for the topsoil (0–30 cm) and subsoil (30–100 cm) at a global scale, with a spatial resolution of 30 arcsec. The resulting maps, along with their associated prediction uncertainties, are freely available on Zenodo.
Overall, I am satisfied with the manuscript and the dataset provided on Zenodo. However, I recommend a few minor revisions before it is suitable for publication. These are as follows:
(1) A few minor spelling and grammatical corrections are needed to enhance the text's readability.
(2) I am curious why the bias (mean error) was not calculated. Demonstrating that the maps are unbiased would strengthen the validity of the results.
(3) I am also wondering why the R-squared value was used instead of the model efficiency coefficient (MEC) (https://en.wikipedia.org/wiki/Nash%E2%80%93Sutcliffe_model_efficiency_coefficient). MEC is often considered more appropriate than R-squared in similar contexts.
(4) Clarification of Section 2.4.2 (Quantification of uncertainty): I believe this section could be revised to provide clearer and more reproducible methodological details. If I understand correctly, did you fit a single QRF model using approximately 90,000 × 100 carbon turnover observations to account for input uncertainties? Or did you fit 100 different QRF models and then combine the results? Clarifying this would enhance transparency.
(5) It would be beneficial to include guidance on how to interpret the PIR values. This would help readers better assess the quality of the maps and understand the magnitude of the prediction uncertainty.
(6) Similarly, a more detailed explanation of PICP would be helpful. Many readers may be unfamiliar with this metric, including what it measures, how to interpret it, and how it reflects the fairness of the uncertainty quantifications.
(7) The sections on mapping and uncertainty quantification appear somewhat intertwined. Did you first fit an RF model to create the maps and then use a separate QRF model (including the inputs’ uncertainty) to quantify the uncertainty associated with the RF predictions? Did you fix the (pseudo-)random number generator to maintain consistency between the RF and QRF models? Clarifying these aspects would enhance the transparency and reproducibility of the methodology.
Once these minor issues are addressed, I believe the manuscript will be ready for publication.
Citation: https://doi.org/10.5194/essd-2024-618-RC3 - AC3: 'Reply on RC3', Lei Zhang, 14 Mar 2025
Status: closed
-
RC1: 'Comment on essd-2024-618', Anonymous Referee #1, 03 Feb 2025
This study creates global maps of soil organic carbon turnover times (τ) at both topsoil and subsoil depths with observation-based datasets and machine learning methods. It addresses an important knowledge gap regarding the global maps related to the depth-dependent variations and environmental controls of soil carbon turnover, which is essential for understanding terrestrial carbon storage and dynamics, especially in the context of climate change. The methods used are well-suited to the scale and complexity of the question being addressed. The use of over ninety thousand geo-referenced soil profiles, satellite-derived environmental variables, and machine learning models, provides a robust framework for mapping and analyzing τ across a wide range of environmental conditions and different biomes. The authors’ approach to quantify the uncertainty of their maps also adds considerable value, making their results more applicable to future carbon cycle modelling and land management efforts. In conclusion, I recommend the acceptance of this manuscript after revisions following several suggestions I listed below.
(1) Organic carbon turnover times can be calculated from influencing factors, such as carbon allocation belowground, SOC stocks. Why not first generate the spatial distribution data of those factors based on sampling data with machine learning method, then calculate organic carbon turnover times with the physical equations.
(2) From the physical equations 1 to 8, the importance of different factors could be derived. Do the importance results of RF coincide with those equations?
(3) While the methods section is thorough, some readers may benefit from further clarification on the choice of specific methods, particularly regarding the machine learning model calibration. A more explicit discussion of how the model’s hyper-parameters were tuned would help readers better appreciate the methodology’s rigor.
(4) The cross-validation procedure is well-executed, but it would be helpful to provide more specific details on how the biome-specific samples were handled during the 10-fold validation process. Ensuring that the sample division properly accounts for the geographical and ecological variation across biomes could further strengthen the model’s credibility.
(5) The paper requires careful attention to language for clarity and readability. I would recommend reviewing some descriptions for possible simplifications in sentence structure and corrections in grammar. Some examples are listed below.
- 18: “… is central to …” -> “… plays a crucial role in …”.
- 19: “remain” -> “remains”.
- 92: it is better to change the word “minimize” to “reduce”.
- 103: “layers observations” -> “layer observations”.
- 465: “insights of this study -> “insights from this study”.
Citation: https://doi.org/10.5194/essd-2024-618-RC1 - AC1: 'Reply on RC1', Lei Zhang, 14 Mar 2025
-
RC2: 'Comment on essd-2024-618', Anonymous Referee #2, 22 Feb 2025
General comments:
This manuscript calculated the apparent turnover time of top and subsoil SOC on a global scale. The major outcome from this work is very useful and timely needed for soil biogeochemistry and carbon cycle modeling communities. The comprehensive data inputs authors used, random forest based geospatial predictive mapping and in-depth uncertainty analysis guaranteed the quality of the produced τ map. Overall, this is a solid and interesting work and I would like to recommend it for acceptance after some technical revisions.
Specific comments:
- The description of quantile regression forest is not very straightforward. Can authors double check and revise this part?
- There lacks certain discussion of the topographical effect on τ. Since the topography largely impacts τ in tundra covered regions (Fig. S23, S24), can authors discuss this finding?
Technical corrections:
I suggest authors double check and add units to all variables in your equations.
Line 89: "collected form" shall be "collected from"
Line 99 - 100: In equation 1, SOC_Du-Dl is not defined.
Line 100: "Were" shall be "Where"
Line 103: "fit layers observations at different depth intervals" can be simplified "fit observations along depth"?
Line 105: "the SOCS at two layers for" shall be more specific "the SOCS at top- and subsoil layers for"
Line 112: "The flux of carbon at a certain soil layer needs to be obtained through" can be revised "Carbon influx at each soil layer comes from"
Line 114: "The annual NPP (kg C m-2 yr-1) produced by the moderate-resolution imaging spectroradiometer (MODIS)" means the MOD17 products? Please add the specific version of the MOD17 product and the url where the author downloaded this data.
Line 126: "total amount of roots" shall be "total root biomass"?
Line 132: " the root distribution" shall be " the root biomass distribution"?
Line 144: " for each soil sample site, root profile observations within the same terrestrial ecoregions (Dinerstein et al., 2017) and the same soil type (FAO–Unesco, 1990) as that of the soil sample were selected. The corresponding mean of those selected root observations for each soil sample location were finally collected (Figs. S8 and S9)" can be simplified "we apply the arithmetic mean of fr from root profile observations within the same terrestrial ecoregions (Dinerstein et al., 2017) and soil type (FAO–Unesco, 1990) as soil sample."
Line 189: "The partial correlation of each influencing factor was calculated while controlling other factors" at which level? Mean or median?
Line 201: "and this division was performed on each biome data to ensure that the ten split sets can keep a balance among biomes". If my understanding is correct, do you mean "and samples of each biome in each subset has the same proportion as the whole dataset"?
Line 209: "by replacing observations by indicator transforms". Not sure I understand "indicator transforms". Do you mean replacing the actual observation values by a tag of category? If so I'm not sure why this is needed. Can you provide a simple example?
Line 213: "has been also" shall be "has also been"
Line 215: If my understanding is correct, it's nice to consider error propagation and produce a dataset considering this uncertainty information to train quantile regression forests. But in your writing this information is not explicitly conveyed so I feel a bit confused when reading this part. It will be better if authors can explicitly tell readers this information above this paragraph.
Line 250: Unfinished sentence "The shaded grey area represents the"
Figure 6: is interesting. I guess the y axis title "MAT/MAP effect" means the partial regression coefficient between MAT/MAP and τ? Please add a description in figure caption.
Figure 6: Another question. I know to completely decorrelate climate and edaphic variables is extremely tough, but I would suggest authors provide a correlation matrix to visualize and identify potential issues with "multicollinearity" between 2 climate and 4 edaphic variables.
Figure 6d shows different responses for topsoil and subsoil. As high CEC enhances adsorption, more nutrients would be available for micro-organism and more mineral-organic compounds are formed, which tends to increase sensitivity of SOCS to temperature. But meanwhile, plants are more productive and may have higher plant carbon stock, thus might increase sensitivity of NPP to temperature. The results might be the combination of both effects.
Figure 7c: What is the unit of variable importance (%)?
Line 405: "biogeochemical simulations by ESMs, and will be useful to improve" can be simplified to "ESM simulations and improve"
Line 407: shall mention CMIP6 outputs from which experiment?
From Fig. S17, it seems like most of the wetland samples are collected from the tropics. Would you justify the performance of wetland τ the arctic region?
Citation: https://doi.org/10.5194/essd-2024-618-RC2 - AC2: 'Reply on RC2', Lei Zhang, 14 Mar 2025
-
RC3: 'Comment on essd-2024-618', Anonymous Referee #3, 05 Mar 2025
The manuscript titled “Mapping global distributions, environmental controls, and uncertainties of apparent top- and subsoil organic carbon turnover times” addresses an important knowledge gap and is therefore likely to be of broad interest. The primary aim of the study is to map carbon turnover times for the topsoil (0–30 cm) and subsoil (30–100 cm) at a global scale, with a spatial resolution of 30 arcsec. The resulting maps, along with their associated prediction uncertainties, are freely available on Zenodo.
Overall, I am satisfied with the manuscript and the dataset provided on Zenodo. However, I recommend a few minor revisions before it is suitable for publication. These are as follows:
(1) A few minor spelling and grammatical corrections are needed to enhance the text's readability.
(2) I am curious why the bias (mean error) was not calculated. Demonstrating that the maps are unbiased would strengthen the validity of the results.
(3) I am also wondering why the R-squared value was used instead of the model efficiency coefficient (MEC) (https://en.wikipedia.org/wiki/Nash%E2%80%93Sutcliffe_model_efficiency_coefficient). MEC is often considered more appropriate than R-squared in similar contexts.
(4) Clarification of Section 2.4.2 (Quantification of uncertainty): I believe this section could be revised to provide clearer and more reproducible methodological details. If I understand correctly, did you fit a single QRF model using approximately 90,000 × 100 carbon turnover observations to account for input uncertainties? Or did you fit 100 different QRF models and then combine the results? Clarifying this would enhance transparency.
(5) It would be beneficial to include guidance on how to interpret the PIR values. This would help readers better assess the quality of the maps and understand the magnitude of the prediction uncertainty.
(6) Similarly, a more detailed explanation of PICP would be helpful. Many readers may be unfamiliar with this metric, including what it measures, how to interpret it, and how it reflects the fairness of the uncertainty quantifications.
(7) The sections on mapping and uncertainty quantification appear somewhat intertwined. Did you first fit an RF model to create the maps and then use a separate QRF model (including the inputs’ uncertainty) to quantify the uncertainty associated with the RF predictions? Did you fix the (pseudo-)random number generator to maintain consistency between the RF and QRF models? Clarifying these aspects would enhance the transparency and reproducibility of the methodology.
Once these minor issues are addressed, I believe the manuscript will be ready for publication.
Citation: https://doi.org/10.5194/essd-2024-618-RC3 - AC3: 'Reply on RC3', Lei Zhang, 14 Mar 2025
Data sets
Global maps of top- and subsoil organic carbon turnover times Lei Zhang, Lin Yang, Thomas W. Crowther, Constantin M. Zohner, Sebastian Doetterl, Gerard B. M. Heuvelink, Alexandre M. J.-C. Wadoux, A.-Xing Zhu, Yue Pu, Feixue Shen, Haozhi Ma, Yibiao Zou, and Chenghu Zhou https://doi.org/10.5281/zenodo.14560239
Model code and software
Code for the analyses of global top- and subsoil organic carbon turnover times Lei Zhang https://github.com/leizhang-geo/global_soil_carbon_turnover_time.git
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 544 | 135 | 11 | 690 | 31 | 7 | 7 |
- HTML: 544
- PDF: 135
- XML: 11
- Total: 690
- Supplement: 31
- BibTeX: 7
- EndNote: 7
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1