the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Global patterns and drivers of soil dissolved organic carbon concentrations
Abstract. Dissolved organic carbon (DOC) is the most active carbon pool in soils, which plays critical roles in soil carbon cycling, plant productivity, and global climate change. An accurate assessment of the quantity of DOC in the soil is essential for the detailed elucidation of ecosystem functions and services. Nevertheless, the global driving factors and distribution of soil DOC remain inadequately quantified due to the scarcity of large-scale data. Here, a comprehensive global database of 12807 soil DOC concentrations derived from 975 target papers in the literature was compiled. Detailed geographic locations, climate, and soil properties were also recorded as predictors of soil DOC. Machine learning techniques were employed to assess the relative importance of various predictors in the determination of soil DOC concentrations, which were subsequently extended for their prediction on a global scale. The worldwide soil DOC concentration spanned a wide range (0.04 to 7859 mg kg-1), averaging 222.78 mg kg-1. The 12 selected variables (including soil properties, month, climate, and ecosystem) explained 65 % of the variance in soil DOC concentrations. Elevation, soil clay, and soil organic carbon were three of the most important predictors. Global soil DOC concentration increased from the equator to the poles. The soil DOC stocks in the topsoil layer (0–30 cm) amounted to 12.17 Pg, with significant variations observed across different continents. These results are instrumental for informing strategies on soil management practices, ecosystem services, and the mitigation of climate change. Furthermore, our database can be combined with other carbon pools to explore the total soil carbon turnover and constrain Earth carbon models. The dataset is publicly available at https://doi.org/10.6084/m9.figshare.26379898 (Ren and Cai, 2024).
- Preprint
(1072 KB) - Metadata XML
-
Supplement
(2329 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
RC1: 'Comment on essd-2024-343', Anonymous Referee #1, 10 Oct 2024
In this study, the authors produced a global map for soil dissolved organic carbon (DOC) concentration and attributed the global variation to environmental factors. It is an interesting research endeavor, and relevant to the scope of this journal. However, I have a few major concerns.
1. DOC can fluctuate significantly over time. For example, one may see high concentration in growing season and low concentration in non growing season. As a result, the global map produced in this study may contain considerable uncertainties due to the influence of temporal variation.
2. Methods, especially global mapping and calculation are too simplified and even missing. Data leakage due to training and validation may inflate the RF model performance. (See details below).
3. The language requires significant improvement.
L123: Elevation does not belong to climate, but topography.
L128-153: While testing a model, please split the data into training, validation, and testing datasets. Data leakage exists without testing datasets. Therefore, your results may be over-confident.
L147-153: move this part to results section for model performance evaluation. Please also expand it to include more information.
L188-190: not clear what the authors are trying to say
L195: it is not normal distribution because the values are natural logged.
L198-200: please provide median in addition to mean.
L202-212: RF model performance may be inflated due to data leakage during model training. The authors need to set aside a testing dataset in addition to training and validation datasets.
L214-226: Month and depth are predictors in the RF model. What values did the “Month” and “depth” take while extrapolating RF to the whole globe? Please also use medians instead of mean to describe DOC concentration due to its strong skew from normal distribution. How did the authors derive DOC in Pg from mg/kg?
L232: please compare median.
Citation: https://doi.org/10.5194/essd-2024-343-RC1 -
RC2: 'Comment on essd-2024-343', Anonymous Referee #2, 18 Nov 2024
The study aims to understand the global distribution and driving factors of soil dissolved organic carbon (DOC) concentrations. objectives include Identifying global patterns of soil DOC concentrations, determining primary factors controlling soil DOC concentrations globally, and quantifying global soil DOC storage. This work is highly appropriate for Earth System Science Data as it presents a valuable new global dataset with clear methodology and thorough documentation.
Novel contribution:
- A comprehensive global soil DOC database to date (12,807 observations from 975 publications)
- Application of machine learning to identify key drivers and predict global patterns
- More accurate estimation of global soil DOC stocks compared to previous studies
Comments:
Abstract
- Add brief mention of validation metrics for the machine learning model
- Include the temporal range of the compiled data
- Specify the spatial resolution of the global predictions
L19: Specify the time period over which these samples were collected
L21: After "Machine learning techniques", specify which ones were used
Introduction
- Consider adding a brief discussion of temporal variations in DOC
- Include more recent references (post-2020) on global carbon cycling
- Expand on the limitations of previous global DOC mapping efforts
L36: Update IPCC citation to most recent report
L54-55: Add recent examples of "extensive research"
Methods
- Data quality:
- Need clearer criteria for handling outliers
- Should explain how temporal variations were addressed
- More detail needed on handling missing predictor variables
- Methodological rigorous:
- Consider using ensemble methods beyond Random Forest
- Add cross-validation across different ecosystems
- Include uncertainty analysis for data standardization process
L91: Add justification for each inclusion criterion
L102: Explain how missing data quality was assessed
L140-141: Specify cross-validation fold number
L150-169: Consider adding a flowchart for model selection process
L170-174: Clarify uncertainty calculation method
Results
- Data presentation:
- Include residual plots for model validation
- Provide more detailed ecosystem-specific analyses
- Additional analyses
- Interaction effects between key drivers
- Sensitivity analysis of model predictions
Discussion
- Expand on implications for carbon cycling models
- Discuss potential impacts of climate change on DOC patterns
- Add recommendations for future sampling efforts
L229-233: Update comparisons with more recent studies
L235-237: Expand on ecosystem differences explanation
L250-254: Add mechanism explanations
L278-299: Consider climate change implications
Figure 1: add sample size for each ecosystem type
Citation: https://doi.org/10.5194/essd-2024-343-RC2
Data sets
Global patterns and drivers of soil dissolved organic carbon concentrations Andong Cai https://doi.org/10.6084/m9.figshare.26379898
Viewed
HTML | XML | Total | Supplement | BibTeX | EndNote | |
---|---|---|---|---|---|---|
392 | 87 | 9 | 488 | 50 | 4 | 7 |
- HTML: 392
- PDF: 87
- XML: 9
- Total: 488
- Supplement: 50
- BibTeX: 4
- EndNote: 7
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1