Deriving a Transformation Rate Map of Dissolved Organic Carbon over the Contiguous U.S.

Li, Lingbo; Li, Hong-Yi; Abeshu, Guta; Tang, Jinyun; Leung, L. Ruby; Liao, Chang; Tan, Zeli; Tian, Hanqin; Thornton, Peter; Yang, Xiaojuan

doi:https://doi.org/10.5194/essd-2024-43

Preprints

https://doi.org/10.5194/essd-2024-43

Preprints

02 Apr 2024

| 02 Apr 2024

Status: this preprint is currently under review for the journal ESSD.

Deriving a Transformation Rate Map of Dissolved Organic Carbon over the Contiguous U.S.

Lingbo Li, Hong-Yi Li, Guta Abeshu, Jinyun Tang, L. Ruby Leung, Chang Liao, Zeli Tan, Hanqin Tian, Peter Thornton, and Xiaojuan Yang

Abstract. Riverine dissolved organic carbon (DOC) plays a vital role in regional and global carbon cycles. However, the processes of DOC conversion from soil organic carbon (SOC) and leaching into rivers are insufficiently understood, inconsistently represented, and poorly parameterized, particularly in land surface and earth system models. As a first attempt to fill this gap, we propose a generic formula that directly connects SOC concentration with DOC concentration in headwater streams, where a single parameter, the transformation rate from SOC in the soil to DOC leaching flux, P_r, accounts for the overall processes governing SOC conversion to DOC and leaching from soils (along with runoff) into headwater streams. We then derive a high-resolution P_r map over the contiguous U.S. (CONUS) in five major steps: 1) selecting 2595 headwater catchments where observed riverine DOC data are available with reasonable quality; 2) estimating catchment-average SOC for the 2595 catchments based on high-resolution SOC data; 3) estimating the P_r values for these catchments based on the generic formula and catchment-average SOC; 4) developing a predictive model of P_r with machine learning (ML) techniques and catchment-scale climate, hydrology, geology, and other attributes; and 5) deriving a national map of P_r, based on the ML model. For evaluation, we compare the DOC concentration derived using the P_r map and the observed DOC concentration values at another 3210 headwater gauges. The resulting mean absolute scaled error and coefficient of determination are 0.73 and 0.47, respectively, suggesting the effectiveness of the overall methodology. Efforts to constrain uncertainty and evaluate the sensitivity of P_r to different factors are discussed. To illustrate the use of such a map, we derive a riverine DOC concentration reanalysis dataset for more than two million small catchments over CONUS. The map, robustly derived and empirically validated, lays a critical cornerstone for better simulating the terrestrial carbon cycle in land surface and earth system models. Our findings not only set a foundation for improving our predictive understanding of the terrestrial carbon cycle at the regional and global scales but also hold promises for informing policy decisions related to decarbonization and climate change mitigation.

Received: 04 Feb 2024 – Discussion started: 02 Apr 2024

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this preprint. The responsibility to include appropriate place names lies with the authors.

Download & links

Preprint (PDF, 2118 KB)

Supplement (1675 KB)

Download & links

Lingbo Li, Hong-Yi Li, Guta Abeshu, Jinyun Tang, L. Ruby Leung, Chang Liao, Zeli Tan, Hanqin Tian, Peter Thornton, and Xiaojuan Yang

Status: final response (author comments only)

RC1:
'Comment on essd-2024-43', Chuanqi He, 21 May 2024

Review report for ESSD-2024-43
'Deriving a Transformation Rate Map of Dissolved Organic Carbon over the Contiguous U.S.'
This study presents an innovative approach to deriving a high-resolution transformation rate (Pr) map of dissolved organic carbon (DOC) from soil organic carbon (SOC) across the contiguous United States using machine learning techniques, specifically XGBoost. By predicting DOC transformation rate based on various environmental attributes, the authors generate a DOC concentration reanalysis dataset for over two million small catchments. This research addresses the insufficient understanding of DOC conversion processes and provides a robust methodology for improving carbon cycle simulations in land surface and earth system models. The use of XGBoost to derive Pr values and create a high-resolution DOC concentration map is both novel and effective. The study's comprehensive data analysis and rigorous machine learning framework ensure robust and reliable results. Additionally, the findings have significant implications for enhancing carbon cycle models and informing climate change mitigation policies.
       The article is logically structured, and the objectives are clear. The introduction is well-written and accessible, even to a geomorphologist like myself. According to the link provided in the 'Data availability' section, I downloaded all the Zenodo data. I opened and checked the raw data in ArcGIS, which includes SOC, PR, and DOC for the United States. The authors also provided a readme file explaining the attribute tables and their units.
       If the authors can address my comments below and those of other reviewers through a major revision, the manuscript can be a valuable contribution to ESSD.

Comments for manuscript:
This database provides an excellent method for calculating DOC. However, it is limited to the United States. If other scholars wish to apply this method to other regions, the contribution of this paper would be even greater. Although the authors mention in the first paragraph of the Method section that 'The methodology here is described with specific details over the CONUS region, but it is transferable to other regions after some modifications based on data availability', they do not provide further details on how to apply this method to other regions. Especially since some environmental factors have been found to have significant impacts while others do not. This information is crucial for applications in other regions. It is suggested that the authors discuss this briefly in the Potential use section (how the methods and results of this paper can inspire the calculation of DOC or Pr in other regions globally).

The article is filled with numerous abbreviations, including those in the main text, figures, and tables, which increases the difficulty for readers. It is recommended that the authors avoid retaining so many abbreviations, especially those that appear infrequently in the later sections (e.g., fewer than five times) and those that are not used at all later on, for example, abbreviations like 'Pg' and 'ESMs' in the Introduction. In Tables 1, 2, 4, and 5, as a reader, I cannot understand what these parameters represent just by looking at the tables. Moreover, the parameter names in the 'Attributes' column of Table 1 are entirely unclear. Recommendations: Avoid using abbreviations in the tables. If abbreviations must be used, provide explanations at the bottom of the table to help readers understand, rather than having them search the main text or supplementary information for the full terms. The content in the 'Attributes' column of Table 1 is completely incomprehensible. It is suggested to move Table 1 to the supplementary information.

The authors selected headwaters based on the following two criteria: 1) there are no upstream rivers flowing into them, and 2) their drainage areas are no more than 2500 km². I have the following questions:
1）Does the first criterion mean that the selected station can only have one river upstream without any tributaries?
2）If so, the second criterion excludes drainage areas larger than 2500 km². I find it hard to believe that a watershed of several thousand square kilometers has only one river without any tributaries. Please provide more details in the main text to clarify this and avoid confusion for readers like me.
3）Additionally, according to Fig. S1, the minimum drainage area in NHDPlus is 0.001 km², approximately a 30-meter square. As a geomorphologist, I do not understand how a watershed of this size can have sufficient upstream drainage area to form a river. Generally, river sources are not at the drainage divide but are 500-5000 meters downstream from it.
4）Moreover, I hope the authors clarify in the caption of Fig. S1 whether the data source includes all watersheds in NHDPlus or only the several thousand watersheds used in this study.

The authors need to explain in the main text the format and size of all the files uploaded to Zenodo, especially detailing what information is included in the files with the format 'gpkg' and what software readers can use to open and edit them.

Among the 29,320 WQP stations, some stations do not have existing upstream watershed boundaries. In such cases, the authors obtained the watershed boundaries using DEM. I have the following questions regarding this:
1) The authors need to clarify, among the 22,201 stations, how many watershed boundaries were derived from DEM and how many from NHDPlus?
2) What resolution of DEM was used, and how were the watershed boundaries calculated?
3) The authors should compare their calculated watershed boundaries with the global watershed boundaries based on a 90-m resolution DEM and advanced algorithms (ESSD 6, 1151–1166, 2024) and present a comparison figure in the SI.

The authors should provide a clear definition of "headwater" as used in this paper in the introduction. Is it determined based on stream order, river length, drainage area, or the number of tributaries? Additionally, they need to explain why they focus on headwaters.

Maybe convert Table 2 into a bar chart and place it in SI. Additionally, again, many abbreviations in Table 2 need to be explained with their full terms.

The language in this article needs further refinement. Here are just some examples that need to be revised, and the authors should check the entire text:
1) Delete "quickly" from line 149.
2) Delete "required for this study" from line 153.
3) Refine "We collect a wide range of environmental variables, comprising a total of 126 variables" to "We collect 126 environmental variables.".
4) Change "The ML technique used in this study is the eXtreme Gradient Boosting (XGBoost) algorithm" to "We use the eXtreme Gradient Boosting (XGBoost) ML algorithm."

The title is a bit long; it is recommended to change it to: "U.S. Transformation Rate Map of Dissolved Organic Carbon" or "Transformation Rate Map of Dissolved Organic Carbon in the Contiguous U.S."

The citation format for figures is completely inconsistent throughout the text. Examples for the same figure include: Fig. S1, supplementary Fig. S1, and Supplementary Fig. S1. Please check the entire text (main text, figures, SI) and standardize according to ESSD requirements.

L175 ScienceBase also provides indicators of human activities, right?

L244 "Out of the remaining 95 variables (see supplementary Tables S1 and S2 for details), 46 are relatively independent from each other. However, the other 49 are highly correlated with one or more variables." How did the authors determine "relatively independent" and "highly correlated"? I expect to see more explanation of this in the main text.

Line 249, change "see Supplementary Figure S3" to "Supplementary Figure S3." Please check the entire text for similar instances where "see" is unnecessary.

Line 251: "This new variable is thus independent of the other environmental variables." I do not understand the basis for this statement. Even if the new 9 combined parameters are formed, it is unlikely that they are completely independent of the other 46 parameters. The authors should provide a brief explanation in the main text or delete this sentence.

Lines 273-275 need to be supported by references.

Line 379: "per_canopy" is too difficult to understand.

In some places, it is written as "section," while in others, it is abbreviated as "sect" (e.g., L380).

L413 ‘Note the unit of DOC concentration in water is mostly reported in mg/L (Schelker et al., 2012; Tian et al., 2015b; Langeveld et al., 2020)’. I think this sentence is not important to be in the main text.

L481-482 ’Blue, red, and grey colors are employed to indicate whether dropping the corresponding predictor will result in an increase, decrease, or insignificant change in the model's performance, respectively‘ should be in figure caption, rather than here.

Comments for dataset in Zenodo:
There are many blank "nodata" areas within the CONUS_DOC_MAP, whereas the CONUS_PR_MAP does not have this issue. The authors need to explain this in the main text.

For reproducibility, the authors need to provide the shapefiles (or other similar vector data) for the 2595 watersheds used for machine learning training and the 3210 watersheds used for evaluation, as well as the shapefiles for these 5805 stations. The machine learning codes, as well as the raw data used for training the machine learning model, need to be uploaded to Zenodo; Then provide another link in the manuscript (not https://doi.org/10.5281/zenodo.8339372).

Suggestion for figures:
The background color of all 2D density plots needs to be changed because the background color is included in the color scale. This makes it difficult for readers to distinguish between the data and the background color.

Are the points in Figure 1 outlets or geometric centers of the watersheds? Additionally, it is necessary to indicate in the figure or caption that the gray lines represent rivers and the black lines represent national boundaries. Also, please specify the sources of these two elements.

Figures 4 and 7 contain numerous abbreviations that are not explained in the captions, making it difficult for readers to understand the figures directly.

It is necessary to explain in the caption of Figure 4 what the correlation coefficient is. Is it Spearman rank?

Why are there many nodata areas near the national boundaries in Figure 5?

Fig. S1 needs ticks on the X-axis.

In the main manuscript, I do not understand the differences between the two types of watershed boundaries provided by NHDPlus. Besides, I do not understand Figure S2. It is recommended to use a real terrain example for illustration to show the differences between these two kinds of watershed boundaries. For example, based on Google Earth, mark the river, the two different watersheds, and the DOC station location (outlet).

I don't have research experience with DOC; most of my comments are from a geomorphological perspective, as well as regarding readability and clarity. I hope my suggestions are helpful.
Chuanqi He          MIT, USA   21 May 2024

Citation: https://doi.org/10.5194/essd-2024-43-RC1
- AC2: 'Reply on RC1', Hongyi Li, 25 Jun 2024
  
  We greatly appreciate the thorough and constructive comments from Reviewer 1. Our point-to-point responses are enclosed in the attachment.
  With warm regards
  Hong-Yi Li, on behalf of all-coauthors
  
  Citation: https://doi.org/10.5194/essd-2024-43-AC2
AC1:
'Comment on essd-2024-43', Hongyi Li, 24 Jun 2024

Publisher’s note: this comment is a copy of AC2 and its content was therefore removed.

Citation: https://doi.org/10.5194/essd-2024-43-AC1
- AC3: 'Reply on AC1', Hongyi Li, 14 Aug 2024
  
  We greatly appreciate the positive feedback from Reviewer #2.
  Look forward to the editor's guidance on the next step.
  
  Citation: https://doi.org/10.5194/essd-2024-43-AC3
RC2:
'Comment on essd-2024-43', Anonymous Referee #2, 06 Aug 2024

The paper titled “Deriving a Transformation Rate Map of Dissolved Organic Carbon over the Contiguous U.S.” presents a novel formula that directly links soil organic carbon (SOC) concentration with dissolved organic carbon (DOC) concentration in headwater streams. This formula uses a single parameter, the transformation rate (Pr ), to represent the overall processes governing the conversion of SOC to DOC and its leaching from soils into headwater streams (along with runoff). The authors have developed a high-resolution Pr map for the contiguous U.S. (CONUS), which is robustly derived and empirically validated. This map provides a crucial foundation for improving the simulation of the terrestrial carbon cycle in land surface and Earth system models.
The study is well-organized and well-written, demonstrating high novelty and significantly contributing to riverine DOC modeling. The paper outlines several potential applications of the derived products. Additionally, the authors have thoroughly addressed the uncertainty analysis and limitations of the study. I recommend accepting the manuscript in its current form.

Citation: https://doi.org/10.5194/essd-2024-43-RC2
- AC4: 'Reply on RC2', Hongyi Li, 14 Aug 2024
  
  We greatly appreciate the positive comments from Reviewer #2.
  Look forward to the editor's guidance on the next step.
  
  Citation: https://doi.org/10.5194/essd-2024-43-AC4
RC3:
'Comment on essd-2024-43', Anonymous Referee #3, 28 Sep 2024

The manuscript addresses a significant issue in the field of environmental science, particularly in understanding the dynamics of dissolved organic carbon (DOC) in the context of climate change and carbon cycling. The manuscript makes effective use of the public available datasets, especially the Water Quality Portal (WQP), provide a solid foundation for the analysis. The proposed method of estimating transformation rates from soil organic carbon (SOC) to DOC using a lumped parameter approach is innovative and could simplify large-scale modeling efforts. The model's simplicity and the reduced data requirements are strengths, making it more accessible for application in regions with limited data availability. And lastly, the model's potential to predict riverine DOC concentrations from SOC values is a valuable tool for water quality management and environmental monitoring. However, there are some potential weaknesses for the authors to consider and to improve the quality of the manuscript: (1) Generalizability: The study focuses on the contiguous U.S., and it is unclear how well the findings and models could be generalized to other regions with different environmental conditions. (2) Complexity of DOC Dynamics: The simplification of the model might overlook the complexity of DOC dynamics, including the influence of various biotic and abiotic factors. (3) Validation and Calibration: The manuscript would benefit from a more detailed discussion on the validation and calibration of the model, including the use of independent datasets. (4) Potential Over-simplification: The assumption that riverine DOC degradation in headwater streams is negligible might be an oversimplification, especially in ecosystems with high microbial activity. (5) Lack of Experimental Data: The study relies heavily on existing datasets, and there is a lack of experimental data to support the model's predictions. Overall, the development of a predictive model that can estimate riverine DOC concentrations from SOC values is innovative and has practical applications, I would recommend the manuscript for acceptance with major revision. In addition, I have a few minor comments; please see below:

Line 131: “Eqn. (4) has several advantages” change to “Eqn. (4) has two advantages”.
Line 153: There are much higher spatial resolution SOC data available (e.g. SoilGrids provides 250m resolution data available, see reference below), why chose use HWSD?
Hengl, Tomislav, Jorge Mendes de Jesus, Gerard BM Heuvelink, Maria Ruiperez Gonzalez, Milan Kilibarda, Aleksandar Blagotić, Wei Shangguan et al. "SoilGrids250m: Global gridded soil information based on machine learning." PLoS one 12, no. 2 (2017): e0169748.
Line 219-220: According to the description, 3210 pairs for evaluation are within the catchment of 2595 pairs for ML modeling, therefore they are not independent and the evaluation might biased.
Figure 3: In scatter plots, observed data are typically placed on the y-axis, while simulated data are positioned on the x-axis. I suggest moving the estimated Pr to the y-axis and the simulated Pr to the x-axis. The same recommendation applies to Figure 6.

Citation: https://doi.org/10.5194/essd-2024-43-RC3
- AC5: 'Reply on RC3', Hongyi Li, 31 Oct 2024
  
  We greatly appreciate the insightful comments from Reviewer #3. Our point-to-point responses are enclosed in the attachment.
  With warm regards
  Hong-Yi Li, on behalf of all-coauthors
  
  Citation: https://doi.org/10.5194/essd-2024-43-AC5

Lingbo Li, Hong-Yi Li, Guta Abeshu, Jinyun Tang, L. Ruby Leung, Chang Liao, Zeli Tan, Hanqin Tian, Peter Thornton, and Xiaojuan Yang

Supplement

https://doi.org/10.5194/essd-2024-43-supplement

Data sets

Deriving a Transformation Rate Map of Dissolved Organic Carbon over the Contiguous U.S. Lingbo Li, Hong-Yi Li, Guta Abeshu, Jinyun Tang, L. Ruby Leung, Chang Liao, Zeli Tan, Hanqin Tian, Peter Thornton, and Xiaojuan Yang https://zenodo.org/records/8132850

Lingbo Li, Hong-Yi Li, Guta Abeshu, Jinyun Tang, L. Ruby Leung, Chang Liao, Zeli Tan, Hanqin Tian, Peter Thornton, and Xiaojuan Yang

Viewed

Total article views: 858 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
586	131	141	858	45	44	37

HTML: 586
PDF: 131
XML: 141
Total: 858
Supplement: 45
BibTeX: 44
EndNote: 37

Views and downloads (calculated since 02 Apr 2024)

Month	HTML	PDF	XML	Total
Apr 2024	167	29	18	214
May 2024	102	23	12	137
Jun 2024	78	16	8	102
Jul 2024	35	12	4	51
Aug 2024	94	24	15	133
Sep 2024	37	12	1	50
Oct 2024	39	10	33	82
Nov 2024	34	5	50	89

Cumulative views and downloads (calculated since 02 Apr 2024)

Month	HTML	PDF	XML	Total
Apr 2024	167	29	18	214
May 2024	102	23	12	137
Jun 2024	78	16	8	102
Jul 2024	35	12	4	51
Aug 2024	94	24	15	133
Sep 2024	37	12	1	50
Oct 2024	39	10	33	82
Nov 2024	34	5	50	89

Viewed (geographical distribution)

Total article views: 879 (including HTML, PDF, and XML) Thereof 879 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 20 Nov 2024

Download

Preprint (2118 KB)
Metadata XML

Short summary

We have developed a new map that reveals how organic carbon from soil leaches into headwater streams over the contiguous United States. We use advanced artificial intelligence techniques and a massive amount of data, including observations at over 2,500 gauges and a wealth of climate and environmental information. The map is a critical step in understanding and predicting how carbon moves through our environment, hence a useful tool for tackling climate challenges.


Total:	0
HTML:	0
PDF:	0
XML:	0