Improved understanding of nitrate trends, eutrophication indicators, and risk areas using machine learning

Banerjee, Deep S.; Skákala, Jozef

doi:https://doi.org/10.5194/bg-22-3769-2025

Articles | Volume 22, issue 15

https://doi.org/10.5194/bg-22-3769-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/bg-22-3769-2025

© Author(s) 2025. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 22, issue 15

Research article

|

05 Aug 2025

Research article |

| 05 Aug 2025

Improved understanding of nitrate trends, eutrophication indicators, and risk areas using machine learning

Deep S. Banerjee and Jozef Skákala

Download

Final revised paper (published on 05 Aug 2025)
Supplement to the final revised paper
Preprint (discussion started on 19 Jun 2024)

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2024-1367', Anonymous Referee #1, 11 Aug 2024

Comments on the manuscript are provided in the annotated pdf together with comments on spelling/grammar.
Comments against the review criteria are provided below:
1. Does the paper address relevant scientific questions within the scope of BG?
and
2. Does the paper present novel concepts, ideas, tools, or data?
Yes, the paper addresses scientific questions within the scope of BG. The paper presents a machine learning (ML) approach to generate a surface nitrate concentration dataset for the North West European Shelf (NWES) between 1998 and 2020 and uses publicly available data to test the model performance. The authors use this dataset to evaluate changes in nitrate concentration over this time and its applicability to assess indicators of eutrophication.
3. Are substantial conclusions reached?
The authors present an analysis of their reconstructed dataset to evaluate changes in nitrate concentration between 1998 and 2020 and test its applicability to indicators of eutrophication. They propose further work, including applying similar methods to produce datasets for other biogeochemical variables which would be useful for further testing the applicability to support eutrophication assessments.
4. Are the scientific methods and assumptions valid and clearly outlined?
The scientific methods are clearly presented and the use of a less complex neural network model (not time-lagged) explained and justified.
5. Are the results sufficient to support the interpretations and conclusions?
The validation of the model against independent in situ datasets is clearly presented in terms of model skill.
6. Is the description of experiments and calculations sufficiently complete and precise to allow their reproduction by fellow scientists (traceability of results)?
A link to the ML code is given (https://github.com/neccton-algo) but it is not clear here what the relevant code is on this site - need to clarify.
7. Do the authors give proper credit to related work and clearly indicate their own new/original contribution?
Yes, the authors credit authors for other products and datasets used for developing and for comparing with their model outputs.
8. Does the title clearly reflect the contents of the paper?
The consideration of eutrophication in the paper is fairly light-touch - the focus is on modelled surface nitrate concentrations, their relationship with chlorophyll at large spatial scales and trends in nutrient concentrations rather than considering trends in eutrophication per se. The paper does not consider other important indicators of eutrophication, including dissolved oxygen or changes in phytoplankton species, necessary to consider a more comprehensive analysis of eutrophication. Suggest 'Improved understanding of nutrient trends and eutrophication indicators using machine learning'.
9. Does the abstract provide a concise and complete summary?
Yes, some minor comments in the annotated pdf.
10. Is the overall presentation well structured and clear?
Yes
11. Is the language fluent and precise?
The language is largely fluent, a few minor grammar corrections made in the attached pdf.
12. Are mathematical formulae, symbols, abbreviations, and units correctly defined and used?
Equations for deriving the skill metrics to assess the model are provided and explained.
13. Should any parts of the paper (text, formulae, figures, tables) be clarified, reduced, combined, or eliminated?
Some minor clarifications sought - details in the annotated pdf.
14. Are the number and quality of references appropriate?
Yes
15. Is the amount and quality of supplementary material appropriate?
Yes, the information provided in the supplementary material is necessary to support the model validation. It is appropriately referenced in the main text.

Citation: https://doi.org/10.5194/egusphere-2024-1367-RC1
- AC1: 'Reply on RC1', Deep S. Banerjee, 06 Sep 2024
  
  Dear Reviewer,
  We sincerely thank the reviewer for their valuable comments, suggestions and positive feedback. We feel delighted that they found our work novel and appreciate that it addresses relevant questions from BGC’s point of view. We have made all the minor corrections requested by the reviewer. We list all the reviewer comments below in italics, with our responses detailing the changes in standard fonts:
  
  (1) Section 1 lines 16-17: Reviewer: “Nitrogen is one of the most important components of organic matter' Clarify this statement - in what context? needed for what - primary production?”
  We thank the reviewer for the comment. We have modified the sentence (see the line 16-17, page 1) to
  “Nitrogen is one of the most important components of organic matter, needed for primary production relatively large concentrations, as demonstrated by the Redfield ratios (Tett et al., 1985).”
  
  (2) Section 1 line 28-30: Reviewer: “Increased nutrient pollution may be linked to development of HABs”
  We thank the reviewer, we have now mentioned HAB’s explicitly, see the line 28, page 2:
  “Furthermore, eutrophication events are often dominated by species that produce toxins (i.e. harmful algae blooms, HABs) that have detrimental effects on the marine ecosystem by causing fish kills, seafood contamination, and even posing risks to human lives (Anderson et al., 2012)."
  
  (3) Section 1 line 31-35: Reviewer: “It isn't necessarily a simple relationship - the availability of light plays a significant role in determining growth. Clarify this statement.”
  We agree with the reviewer that the dynamics is more complicated, so we slightly toned down the sentence, please see line 30, page 2:
  “Additionally high nitrate concentrations may lead under certain circumstances to the excessive production of organic matter, ...”
  
  (4) Section 1 line 31-35: Reviewer: “Need to recognise that eutrophication is defined as undesirable disturbance from excess nutrients and therefore eutrophication monitoring programmes must include indicators of undesirable disturbance as well as nutrients (e.g. chlorophyll, dissolved oxygen, phytoplankton species). Make clear here.”
  We agree with the reviewer, and we changed the sentence to (see line 33, page 2):
  “.. where nitrate monitoring and predicting, along with other indicators (e.g. chlorophyll, dissolved oxygen, phytoplankton species), providing an essential tool informing marine management and policy."
  
  (5) Section 1 line 37: Reviewer: “also Seine, Scheldt, Elbe”
  We thank the reviewer for suggesting other important rivers i.e., associated with the study i.e., Seine, Scheldt and Elbe. These are added to the sentence, please see line 37, page 2.
  
  (6) Section 1 line 45: Reviewer: “EU regulations are not set by OSPAR, need to re-word this sentence. From https://www.ospar.org/work-areas/hasec/eutrophication: The implementation of the Eutrophication Strategy takes place within the framework of the obligations and commitments of Contracting Parties under other international agreements. This includes European Union legislation to reduce nutrient discharges and emissions, for example the Nitrates Directive (91/676/EEC), Urban Waste Water Treatment Directive (91/271/EEC), the Water Framework Directive (2000/60/EEC) and the Marine Strategy Framework Directive (2008/56/EC). OSPAR contracting parties work to implement the PARCOM 88/2 agreement to: - take effective national steps in order to reduce nutrient inputs into areas where these inputs are likely, directly or indirectly, to cause pollution; - aim to achieve a substantial reduction (of the order of 50%) in inputs of phosphorus and nitrogen into these areas between 1985 and 1995, or earlier if possible. The recent analysis in the Quality Status Report shows that the reduction in total nitrogen loads to the OSPAR area is driven primarily by decreases in atmospheric nitrogen rather than reductions in riverine nitrogen loads (https://oap.ospar.org/en/ospar-assessments/quality-status-reports/qsr-2023/indicator-assessments/inputs-nutrients/)”
  
  We sincerely thank the reviewer for making this point, we have changed the wording (please see line 45, page 2):
  “However, EU regulations following OSPAR convention ..."
  
  (7) Section 2.2.1 line 136: Reviewer: The figure number needed to be corrected to S2.
  We corrected the typo here and changed the figure number to S2.
  
  (8) Section 2.2.2 line 148: Reviewer: The reviewer asked about the interpolation method. “how was this done?”
  It was through standard linear interpolation, so we added the word “linear”" to the text, please see line 144, page 6.
  
  (9) Section 2.2.2 line 152:
  We changed the sentence to explain the independent test data from Scottish locations used for validating the model.
  
  (10) Section 3.1 line 190: Reviewer: “This should be qualified - still outperforms in terms of BC-RMSE and R, but not consistently in terms of bias.”
  We thank the reviewer for the comment. we have added the relevant information in the sentence (see line 185, page 8):
  “..but the NN model still outperforms the reanalysis in BC-RMSE and R at each of the locations.”
  
  (11) Section 3.2 line 216: Reviewer: “It would be good to clarify what months have been used in this analysis for 'winter' and whether they are the same as those used in the OSPAR Common Procedure i.e. December, January and February.”
  Yes, the winter months have been defined as per OSPAR Common Procedure, being December, January, February. We have made this now clear, see line 185, page 8:
  “... winter (December-February) nitrate concentration,...”
  
  (12) Section 3.3 line 220: Reviewer: “Looking at Fig 6B, the inter annual nitrate variability seems to be 5 – 10% in the majority of this region.”
  We agree with the reviewer that this statement was inaccurate, we have modified it (see line 221, page 11):
  “...(>5% in the majority of the region, with 10-20% in specific sections) ...”
  
  (13) Figure 5: Reviewer: “Need to label which column of figures is from NN and which is from WOA.”
  We thank the reviewer for pointing this out to us. We added the labels for NN and WOA in both columns for clarity.
  
  (14) Section 3.2 line 236: Reviewer: “Suggest referring to the most recent OSPAR eutrophication assessment for latest status of problem/non problem areas (links below). This assessment used much smaller, more ecologically relevant assessment areas. https://www.frontiersin.org/journals/ocean-sustainability/articles/10.3389/focsu.2023.1253923/full https://oap.ospar.org/en/ospar-assessments/quality-status-reports/qsr-2023/thematic-assessments/eutrophication/”
  We thank the reviewer for these valuable references, we have added them to the Axe et al (2017) reference, i.e. see line 238 and page 13:
  “... (Axe et al., (2017), Devlin et al., (2023a, 2023b)) ..."
  
  (15) Section 3.2 line 245: Reviewer: “The analysis of riverine nutrient inputs (1990 - 2019) under the most recent OSPAR eutrophication assessment shows that reductions in atmospheric N have driven reductions in total N inputs to the OSPAR maritime area, reductions in riverine P loads have been much more significant (Devlin et al., 2023, doi: 10.3389/focsu.2023.1253923)”
  We thank the reviewer for this comment, we have modified the sentence on line 247, page 13 to:
  “These small trends may follow the smaller rates of reduction in the nitrate riverine inputs during the data period (1998-
  2020), compared to their large reduction in the 1980’s and earlier 1990’s (Duarte, 2009; Brockmann et al., 2018; Greenwood
  et al., 2019), although significant reduction in atmospheric nitrogen input has been reported for this period (Devlin et al, 2023a).”
  
  (15) Section 4 line 267: Reviewer: “It is not clear here what the relevant code is on this site.”
  We thank the reviewer for the typo, and we modified the relevant GitHub repository link containing the code used for the experiments.
  
  We also extend our sincere gratitude to the reviewer for pointing out the grammatical errors and typos throughout the entire document, all of which have now been corrected. We would like to thank very much the reviewer for their contribution to the improvement of the manuscript.
  
  Best Wishes,
  Deep Banerjee and Jozef Skakala
  
  Dear Reviewer,
  We sincerely thank the reviewer for their valuable comments, suggestions and positive feedback. We feel delighted that they found our work novel and appreciate that it addresses relevant questions from BGC’s point of view. We have made all the minor corrections requested by the reviewer. All the changes can be found highlighted in bold font in the manuscript. We list all the reviewer comments below in red, with our responses detailing the changes in blue:
  (16) Section 1 lines 16-17: Reviewer: “Nitrogen is one of the most important components of organic matter' Clarify this statement - in what context? needed for what - primary production?”
  
  We thank the reviewer for the comment. We have modified the sentence (see the line 16-17, page 1) to
  “Nitrogen is one of the most important components of organic matter, needed for primary production relatively large concentrations, as demonstrated by the Redfield ratios (Tett et al., 1985).”
  (17) Section 1 line 28-30: Reviewer: “Increased nutrient pollution may be linked to development of HABs”
  We thank the reviewer, we have now mentioned HAB’s explicitly, see the line 28, page 2:
  “Furthermore, eutrophication events are often dominated by species that produce toxins (i.e. harmful algae blooms, HABs) that have detrimental effects on the marine ecosystem by causing fish kills, seafood contamination, and even posing risks to human lives (Anderson et al., 2012)."
  
  (18) Section 1 line 31-35: Reviewer: “It isn't necessarily a simple relationship - the availability of light plays a significant role in determining growth. Clarify this statement.”
  
  We agree with the reviewer that the dynamics is more complicated, so we slightly toned down the sentence, please see line 30, page 2:
  
  “Additionally high nitrate concentrations may lead under certain circumstances to the excessive production of organic matter, ...”
  
  (19) Section 1 line 31-35: Reviewer: “Need to recognise that eutrophication is defined as undesirable disturbance from excess nutrients and therefore eutrophication monitoring programmes must include indicators of undesirable disturbance as well as nutrients (e.g. chlorophyll, dissolved oxygen, phytoplankton species). Make clear here.”
  
  We agree with the reviewer, and we changed the sentence to (see line 33, page 2):
  “.. where nitrate monitoring and predicting, along with other indicators (e.g. chlorophyll, dissolved oxygen, phytoplankton species), providing an essential tool informing marine management and policy."
  (20) Section 1 line 37: Reviewer: “also Seine, Scheldt, Elbe”
  We thank the reviewer for suggesting other important rivers i.e., associated with the study i.e., Seine, Scheldt and Elbe. These are added to the sentence, please see line 37, page 2.
  
  (21) Section 1 line 45: Reviewer: “EU regulations are not set by OSPAR, need to re-word this sentence. From https://www.ospar.org/work-areas/hasec/eutrophication: The implementation of the Eutrophication Strategy takes place within the framework of the obligations and commitments of Contracting Parties under other international agreements. This includes European Union legislation to reduce nutrient discharges and emissions, for example the Nitrates Directive (91/676/EEC), Urban Waste Water Treatment Directive (91/271/EEC), the Water Framework Directive (2000/60/EEC) and the Marine Strategy Framework Directive (2008/56/EC). OSPAR contracting parties work to implement the PARCOM 88/2 agreement to: - take effective national steps in order to reduce nutrient inputs into areas where these inputs are likely, directly or indirectly, to cause pollution; - aim to achieve a substantial reduction (of the order of 50%) in inputs of phosphorus and nitrogen into these areas between 1985 and 1995, or earlier if possible. The recent analysis in the Quality Status Report shows that the reduction in total nitrogen loads to the OSPAR area is driven primarily by decreases in atmospheric nitrogen rather than reductions in riverine nitrogen loads (https://oap.ospar.org/en/ospar-assessments/quality-status-reports/qsr-2023/indicator-assessments/inputs-nutrients/)”
  
  We sincerely thank the reviewer for making this point, we have changed the wording (please see the line 45, page 2):
  “However, EU regulations following OSPAR convention ..."
  
  (22) Section 2.2.1 line 136: Reviewer: The figure number needed to be corrected to S2.
  
  We corrected the typo here and changed the figure number to S2.
  
  (23) Section 2.2.2 line 148: Reviewer: The reviewer asked about the interpolation method. “how was this done?”
  
  It was through standard linear interpolation, so we added the word “linear”" to the text, please see line 144, page 6.
  (24) Section 2.2.2 line 152:
  We changed the sentence to explain the independent test data from Scottish locations used for validating the model.
  
  (25) Section 3.1 line 190: Reviewer: “This should be qualified - still outperforms in terms of BC-RMSE and R, but not consistently in terms of bias.”
  We thank the reviewer for the comment. we have added the relevant information in the sentence (see line 185, page 8):
  “..but the NN model still outperforms the reanalysis in BC-RMSE and R at each of the locations.”
  
  (26) Section 3.2 line 216: Reviewer: “It would be good to clarify what months have been used in this analysis for 'winter' and whether they are the same as those used in the OSPAR Common Procedure i.e. December, January and February.”
  Yes, the winter months have been defined as per OSPAR Common Procedure, being December, January, February. We have made this now clear, see line 185, page 8:
  “... winter (December-February) nitrate concentration,...”
  
  (27) Section 3.3 line 220: Reviewer: “Looking at Fig 6B, the inter annual nitrate variability seems to be 5 – 10% in the majority of this region.”
  
  We agree with the reviewer that this statement was inaccurate, we have modified it (see line 221, page 11):
  “...(>5% in the majority of the region, with 10-20% in specific sections) ...”
  
  (28) Figure 5: Reviewer: “Need to label which column of figures is from NN and which is from WOA.”
  We thank the reviewer for pointing this out to us. We added the labels for NN and WOA in both columns for clarity.
  (29) Section 3.2 line 236: Reviewer: “Suggest referring to the most recent OSPAR eutrophication assessment for latest status of problem/non problem areas (links below). This assessment used much smaller, more ecologically relevant assessment areas. https://www.frontiersin.org/journals/ocean-sustainability/articles/10.3389/focsu.2023.1253923/full https://oap.ospar.org/en/ospar-assessments/quality-status-reports/qsr-2023/thematic-assessments/eutrophication/”
  We thank the reviewer for these valuable references, we have added them to the Axe et al (2017) reference, i.e. see line 238 and page 13:
  “... (Axe et al., (2017), Devlin et al., (2023a, 2023b)) ..."
  
  (15) Section 3.2 line 245: Reviewer: “The analysis of riverine nutrient inputs (1990 - 2019) under the most recent OSPAR eutrophication assessment shows that reductions in atmospheric N have driven reductions in total N inputs to the OSPAR maritime area, reductions in riverine P loads have been much more significant (Devlin et al., 2023, doi: 10.3389/focsu.2023.1253923)”
  
  We thank the reviewer for this comment, we have modified the sentence on the line 247, page 13 to:
  
  “These small trends may follow the smaller rates of reduction in the nitrate riverine inputs during the data period (1998-
  2020), compared to their large reduction in the 1980’s and earlier 1990’s (Duarte, 2009; Brockmann et al., 2018; Greenwood
  et al., 2019), although significant reduction in atmospheric nitrogen input has been reported for this period (Devlin et al, 2023a).”
  
  (30) Section 4 line 267: Reviewer: “It is not clear here what the relevant code is on this site.”
  We thank the reviewer for the typo, and we modified the relevant GitHub repository link containing the concerned code used for the experiments.
  
  We also extend our sincere gratitude to the reviewer for pointing out the grammatical errors and typos throughout the entire document, all of which have now been corrected. We would like to thank very much the reviewer for their contribution to the improvement of the manuscript.
  
  Best Wishes,
  Deep Banerjee and Jozef Skakala
  
  Citation: https://doi.org/10.5194/egusphere-2024-1367-AC1
RC2:
'Comment on egusphere-2024-1367', Anonymous Referee #2, 30 Sep 2024

Please see the attached review report.

Citation: https://doi.org/10.5194/egusphere-2024-1367-RC2
- AC2: 'Reply on RC2', Deep S. Banerjee, 15 Nov 2024
  
  Dear Reviewer,
  We would like to express our sincere gratitude for your valuable comments and detailed review which enabled us to improve our manuscript. Kindly find the detailed response in the attached pdf file.
  Best Wishes,
  Deep S. Banerjee and Jozef Skakala
  
  Citation: https://doi.org/10.5194/egusphere-2024-1367-AC2

Peer review completion

AR: Author's response | RR: Referee report | ED: Editor decision | EF: Editorial file upload

ED: Reconsider after major revisions (17 Dec 2024) by Jamie Shutler

AR by Deep S. Banerjee on behalf of the Authors (16 Jan 2025) Author's response Author's tracked changes Manuscript

ED: Publish subject to minor revisions (review by editor) (21 Feb 2025) by Jamie Shutler

AR by Deep S. Banerjee on behalf of the Authors (25 Feb 2025) Author's response Author's tracked changes Manuscript

ED: Publish as is (17 Mar 2025) by Jamie Shutler

AR by Deep S. Banerjee on behalf of the Authors (24 Mar 2025)

Post-review adjustments

AA: Author's adjustment | EA: Editor approval

AA by Deep S. Banerjee on behalf of the Authors (29 Jul 2025) Author's adjustment Manuscript

EA: Adjustments approved (30 Jul 2025) by Jamie Shutler

Download

Article (8764 KB)
Full-text XML

Short summary

Nitrate is a crucial nutrient in oceans, whose excess can trigger uncontrolled algae growth that damages marine ecosystems. We used machine learning to generate skilled, gap-free, bi-decadal surface nitrate data from sparse observations, revealing areas on the North-West European Shelf that are more vulnerable to excess algae growth if nutrient pollution occurs. We also looked at bi-decadal trends in coastal nitrate and the impact of winter nitrate on spring phytoplankton blooms.