A Probabilistic Approach to Mapping the Contribution of Individual Riverine Discharges into Liverpool Bay Using Distance Accumulation Cost Methods on Satellite Derived Ocean-Colour Data

Round 1

Reviewer 1 Report

This manuscript analyses Dissolved Inorganic Nitrogen (DIN) riverine discharge into Liverpool Bay area. It combines satellite measurements of to measure riverine plumes extent, a surface current measurement model and DIN measurements. Authors implemented a probabilistic approach in a GIS for mapping the spatial-temporal dynamic of DIN in the study area.

The objective of the manuscript is interesting and combines several datasources and techniques to deal with the complex DIN dynamic in the study area. The main limitations of the manuscript are the following:

1) Results validation has not been completed. This problem is persistent through the manuscript. For example, riverine plume extent has been delimited only based on the FUI index. This result could be reinforced by employing water quality parameters (e.g., EC, TSS) that may help to assess the plume extend delimitation. Another example is the absence of validation of DIN concentrations estimated with the probabilistic model. It’s not possible to assess the relative importance of modelled surface currents on DIN predictions or the accuracy of those predictions through the year. This is a critical issue. In any case, a comprehensive limitations of the study section is needed.

2) Information of the water sampling survey is very incomplete (lines 221-223). There is no information about the location, date or depth of sampling, water quality analytical procedures, etc. We need to know when water samples were collected and if water samples have been collected with the correct spatial pattern.

3) A more quantitative analysis of FUI and DIN is needed in order to assess the utility of your remote sensing approach. This is the most important contribution of remote sensing for your study and should be improved.

4) Although the description of the methods has a good structure, some relevant information is missed or may be completed. For example:

• Line 157. Include additional details of the remote sensing dataset (e.g., date of acquisition, spatial resolution, etc.)

• Line 157. Include a reference for POLYMER

• Line 171. Include a reference for the surface currents model.

• Line 175. Include details of employed R packages.

In my opinion, this research is interesting but requires a comprehensive review in order to avoid (or minimize) the limitations of the current manuscript.

 

Author Response

Comments and Suggestions for Authors

This manuscript analyses Dissolved Inorganic Nitrogen (DIN) riverine discharge into Liverpool Bay area. It combines satellite measurements of to measure riverine plumes extent, a surface current measurement model and DIN measurements. Authors implemented a probabilistic approach in a GIS for mapping the spatial-temporal dynamic of DIN in the study area.

The objective of the manuscript is interesting and combines several datasources and techniques to deal with the complex DIN dynamic in the study area. The main limitations of the manuscript are the following:

Authors response: Thank you for reading our manuscript and providing us with many useful comments. We are glad that you found the paper interesting and we have tried to answer your comments below.

1) Results validation has not been completed. This problem is persistent through the manuscript. For example, riverine plume extent has been delimited only based on the FUI index. This result could be reinforced by employing water quality parameters (e.g., EC, TSS) that may help to assess the plume extend delimitation. Another example is the absence of validation of DIN concentrations estimated with the probabilistic model. It’s not possible to assess the relative importance of modelled surface currents on DIN predictions or the accuracy of those predictions through the year. This is a critical issue. In any case, a comprehensive limitations of the study section is needed.

Authors response:

1. Validation of the use of FUI from remote sensing products to define the extent of the river plume has been covered extensively by others (see Fronkova et al 2022, Devlin et al 2012 and references therein). In our study, an almost identical approach to Fronkova et al (2022) was adopted except that the boundary represents the maximum possible extent of the river plume. This was done to confine the analysis area whilst ensuring all of the potential river plume area was included. The aim of this study was not to demark the river plume across the time series. To make this clearer to the reader text has been added – lines 213-214.
2. Limitations in our approach are considered throughout the manuscript however the suggestion of a comprehensive overview is welcome. We have included a paragraph in the discussion section to address these (lines 479 - 495) and have included limitations regarding the use of FUI as a proxy for DIN; a lack of in situ data to enable complete validation; the influence of compounds in the water affecting the optical properties making the approach likely to be regional specific.
3. It may have been unclear in the paper that we did not take water quality samples across the river plume to validate the approach. This would be an extensive sampling approach which is currently not done and hence the reason for the study. We have added text to make this clearer along with better links to the datasets and a map of the locations.

2) Information of the water sampling survey is very incomplete (lines 221-223). There is no information about the location, date or depth of sampling, water quality analytical procedures, etc. We need to know when water samples were collected and if water samples have been collected with the correct spatial pattern.

Authors response: The water quality values were taken from published datasets which was not made clear enough in the text. This has been amended to direct the reader to where the data was obtained (lines 238-244) and we have provided an additional map (Figure S2.1) in the Supplementary materials to show where these data points are situated within the river plume boundary.

3) A more quantitative analysis of FUI and DIN is needed in order to assess the utility of your remote sensing approach. This is the most important contribution of remote sensing for your study and should be improved.

Authors response: The reviewer is pointed to the work of Fronkova et al 2022 which provides the quantitative analysis requested. The paper shows the correlation between FUI and various water quality parameters including nitrate, nitrite and ammonium for the study area. The authors acknowledge that this should have been made clearer and have sought to do so – see lines 238 to 240.

4) Although the description of the methods has a good structure, some relevant information is missed or may be completed. For example:

• Line 157. Include additional details of the remote sensing dataset (e.g., date of acquisition, spatial resolution, etc.)

Authors response: This information is now provided in the manuscript. Lines 163-166.

• Line 157. Include a reference for POLYMER

Authors response: The reference has been included – line 159.

• Line 171. Include a reference for the surface currents model.

Authors response: The reference and location of the surface currents model used in the study is shown in the “Data Availability Statement”. This is made clearer in the text – lines 179.

• Line 175. Include details of employed R packages.

Authors response: The packages used have been included. Apologies for this oversight. Lines 179-180

In my opinion, this research is interesting but requires a comprehensive review in order to avoid (or minimize) the limitations of the current manuscript.

Reviewer 2 Report

The study investigates temporal changes of water quality in the coastal zones. For this aim, Dissolved Inorganic Nitrogen (DIN) is assessed in the study. Liverpool Bay (UK) is selected for the application. The subject is very important and the study is valuable in terms of water quality risk assessment near coastal areas but the novelty of the study is emphasized insufficiently. Some suggestions and comments to the authors are presented below:

1. The flowchart of the suggested methodology should be given by more branches and in detail in Figure 2. Thus, the readers can easily follow the application procedures.

2. The figures should be presented with higher resolution.

3. Conclusions part can be improved in the paper. Here is presented in a general concept.

4. What is the novelty of the paper? The used traditional methods are explained in the paper. Supported and related studies should be strongly presented in the paper by emphasizing the novelty of the paper.

5. Literature part is looking weak. Give main and important examples from literature about “water quality” as

(2010). Identification of coastal water quality by statistical analysis methods in Daya Bay, South China Sea. Marine Pollution Bulletin, 60(6), 852-860.

(2013). Water quality tendency of Akarcay River between 2006–2011. Pamukkale University Journal of Engineering Science, 19(3), 127–132.

 

6. Is the suggested methodology in the paper valid for all areas or is there any limitation or classification for the application?

 

7. The statistical properties as skewness, coefficient of variation, confidence intervals, distribution characteristics, min, max and median etc. of used data as monthly DIN should be given in Table 1, additionally mean.

Check the tenses of the sentences. There are present and present past in a paragraph. See Abstract …

There are some crucial errors.

Keywords should be ordered A to Z.

One sentence can’t be a paragraph. See the lines 264-267 …

 

Use passive sentences. Check the sentences started by “we”.

Author Response

Comments and Suggestions for Authors

The study investigates temporal changes of water quality in the coastal zones. For this aim, Dissolved Inorganic Nitrogen (DIN) is assessed in the study. Liverpool Bay (UK) is selected for the application. The subject is very important and the study is valuable in terms of water quality risk assessment near coastal areas but the novelty of the study is emphasized insufficiently. Some suggestions and comments to the authors are presented below:

Authors response: Thank you for reading our manuscript and providing comments for improvement. We are pleased that you think the study is important and hope that the modifications made have improved the manuscript to your satisfaction. Please find below comments to your specific points made.

1. The flowchart of the suggested methodology should be given by more branches and in detail in Figure 2. Thus, the readers can easily follow the application procedures.

Authors response: The figure is an overview of the process and has been modified to include the relevant methods section for the reader to find out more information. Hopefully this will enable the process to be followed more easily.

2. The figures should be presented with higher resolution.

Authors response: All figures have been provided separately to the journal either as vector graphics or as jpeg files with a resolution of 1200dpi.

3. Conclusions part can be improved in the paper. Here is presented in a general concept.

Authors response: The conclusions section has been modified – see lines 555-573.

4. What is the novelty of the paper? The used traditional methods are explained in the paper. Supported and related studies should be strongly presented in the paper by emphasizing the novelty of the paper.

Authors response: This has been made clearer in the text – see lines 469. The novelty is the combined use of probabilistic approaches for plume presence, potential DIN exposure, and potential riverine input to map contributions from separate rivers contributing to the overall river plume.

5. Literature part is looking weak. Give main and important examples from literature about “water quality” as

(2010). Identification of coastal water quality by statistical analysis methods in Daya Bay, South China Sea. Marine Pollution Bulletin, 60(6), 852-860.

(2013). Water quality tendency of Akarcay River between 2006–2011. Pamukkale University Journal of Engineering Science, 19(3), 127–132.

Authors response: Thank you for directing us to these papers although they are over 10 years old.

The first paper deals with a set of in situ water quality parameters sampled from 12 stations within the study area of Daya Bay. Seasonal samples were taken over a single year (2003) from the surface and bottom water. The results are analysed by cluster analysis/PCA to provide an assessment of trophic status. In the paper there was no use of remote sensing data, and the water quality data was not used to produce a spatial map of water quality across the bay. Cluster analysis was used to infer linkages between sample stations, but the spatial information was limited to 12 point locations, and the temporal data to a daily sample once every quarter. Our study was not to analyse the in situ data but to use this to determine the relationship between the water quality parameter concentration and the FUI value for the same location at the same time.

Unfortunately, the second paper appears to only consider water quality parameters within a river basin and does not involve water quality within transitional or coastal waters (it is also a Turkish language paper).

6. Is the suggested methodology in the paper valid for all areas or is there any limitation or classification for the application?

Authors response: Thank you for the comment. To improve, limitations of the approach are now discussed in a separate paragraph – see lines 479-495.

7. The statistical properties as skewness, coefficient of variation, confidence intervals, distribution characteristics, min, max and median etc. of used data as monthly DIN should be given in Table 1, additionally mean.

Authors response: The table is provided to provide the reader with an indication of the proportional loads of DIN from the 12 rivers. Our study was not to perform statistical analysis on the distributions of these loads across the annual cycle, nor to ascribe any statistical significance. The additional statistical properties can be provided but it makes little sense when they will not be referred to in the text or the analysis.

 

Comments on the Quality of English Language

Check the tenses of the sentences. There are present and present past in a paragraph. See Abstract …

Author’s response: The opening sentences of the abstract have been rewritten to ensure consistency – lines 12-18. Hopefully this improves the text to the satisfaction of the reviewer.

There are some crucial errors.

Author’s response: Unable to comment as these crucial errors are not provided.

Keywords should be ordered A to Z.

Author’s response: Done. Lines 30-31

One sentence can’t be a paragraph. See the lines 264-267 …

Author’s response: This is not true. The definition of a paragraph from Collins is “A paragraph is a section of a piece of writing. A paragraph always begins on a new line and contains at least one sentence.” However, this line has been modified to improve readability.

Use passive sentences. Check the sentences started by “we”.

Author’s response: Many scientific articles involve the active voice (we, our etc) and there are no guidelines for the journal to suggest passive voice must be used. Whilst this style may not be to the liking of the reviewer it does not make the subject matter unreadable.

Reviewer 3 Report

This manuscript is devoted the assessment of dissolved inorganic nitrogen (DIN) in Liverpool Bay (UK) by utilizing a probabilistic approach with satellite-derived ocean-color data. This work is valuable for the assessment of water quality in the coastal zone in remote sensing area. It seems that this manuscript focuses mainly on the technical aspects, while relative few attentions are paid to the theoretical background and related methods.

Specific comments are as follows:

1. It seems that the manuscript focuses mainly on the data processing. It is somewhat difficult for readers to understand the method what the authors described in this paper. Hence, more discussions on models and algorithms of probabilistic approach used in this paper should be added in detail.

2. As the authors stated, this manuscript presented a probabilistic method for assessment of dissolved inorganic nitrogen (DIN) in Liverpool Bay (UK) using distance-accumulation cost methods and an ocean-color product derived from SENTINEL-3 data. I think it is necessary to make a comparison between the probabilistic method and other methods.

Author Response

This manuscript is devoted the assessment of dissolved inorganic nitrogen (DIN) in Liverpool Bay (UK) by utilizing a probabilistic approach with satellite-derived ocean-color data. This work is valuable for the assessment of water quality in the coastal zone in remote sensing area. It seems that this manuscript focuses mainly on the technical aspects, while relative few attentions are paid to the theoretical background and related methods.

Specific comments are as follows:

1. It seems that the manuscript focuses mainly on the data processing. It is somewhat difficult for readers to understand the method what the authors described in this paper. Hence, more discussions on models and algorithms of probabilistic approach used in this paper should be added in detail.

Authors response: The flow diagram in Figure 2 has been updated to include the sections used in generating the probabilities that are used in the approach. In addition, the distance accumulation approach has been covered to guide the reader a little more – lines 198-206. Hopefully this had made it easier to understand the approach.

2. As the authors stated, this manuscript presented a probabilistic method for assessment of dissolved inorganic nitrogen (DIN) in Liverpool Bay (UK) using distance-accumulation cost methods and an ocean-color product derived from SENTINEL-3 data. I think it is necessary to make a comparison between the probabilistic method and other methods.

Authors response: This is the first attempt to use a probabilistic method for mapping DIN risk in UK waters and was necessitated by a lack of other approaches with the same spatial and temporal extent. To our knowledge the only other method to measure this against would be that derived from models. These suffer from a lack of in situ data for nutrient concentrations, especially further out from the coast and can lack accuracy in the near shore waters (see the map in Figure S2.1). Future work will seek to validate the findings although this is not a small task.  Hopefully, the limitations and caveats added to the manuscript will ensure the reader is aware of the current pitfalls of the approach – see lines 479-496 and Conclusions 555-573.

 

Reviewer 4 Report

This paper titled “A probabilistic approach to mapping the contribution of individual riverine discharges into Liverpool Bay using distance accumulation cost methods on satellite derived ocean-colour data” by Heal et al. describes an interesting study using the ArcGIS cost distance function and remote sensing dataset to derive the probabilistic distribution of dissolved inorganic nitrogen (DIN) in the Liverpool bay. It is significant to look at the probable distribution of DIN over the bay, but more importantly, to track the contribution from each of the river sources.  The paper is well written and organized,   and the logic is sound. The study method can be applied to other coastal areas of the world ocean for water quality minitoring. Overall, this is a good paper and fits the scope of this special issue.

I recommend its publication to Remote Sensing with some modification.

Major comments:

1.)    Some context is needed to explain why the cost distance is used. What’s the advantage and how it works in general?   How are the cost surface raster maps with different variables (current diction, FUI values) used in the cost distance calculation?  

2.)    While the probability maps of DIN distribution can be derived, no actual validation is carried out in this study.  How do the authors know their results are realistic?

3.)    How the current velocity/direction affect your results?  Current direction is used to construct cost surface. Why the current speed is not used? How accurate are the reanalysis ocean current datasets?  A discussion on this should be given.  

4.)    Regarding the forces governing the distribution of DIN, what are the roles of the winds, tidal induced mixing, and fresh water transport in this study? The effects of those are not included in the algorithm (at least not in constructing the cost surface).  

Minor Comments:

1.)    Check the figure longitude/latitude labels to make sure they are consistent with different figures and clear visible.  

2.)    Line 541: “normalized cost surfaces can be found in the GitHub repository”, but the git link shows only a README file.

3.)    Some acronyms need to be spell out, such EO at line 124, OSPAR at line 90 etc.

4.)    Line 178, the current is interpolated from model data with nearest neighbor approach. Please give the original reanalysis/model resolution. If the model resolution is coarse, the nearest neighbor method may have issue.   

5.)    Conclusion section needs rewritten; it is more like still in discussion.

Author Response

This paper titled “A probabilistic approach to mapping the contribution of individual riverine discharges into Liverpool Bay using distance accumulation cost methods on satellite derived ocean-colour data” by Heal et al. describes an interesting study using the ArcGIS cost distance function and remote sensing dataset to derive the probabilistic distribution of dissolved inorganic nitrogen (DIN) in the Liverpool bay. It is significant to look at the probable distribution of DIN over the bay, but more importantly, to track the contribution from each of the river sources.  The paper is well written and organized,   and the logic is sound. The study method can be applied to other coastal areas of the world ocean for water quality minitoring. Overall, this is a good paper and fits the scope of this special issue.

I recommend its publication to Remote Sensing with some modification.

Authors response: Thank you for reading our manuscript and we are delighted that you consider the paper to be good. Where possible we have taken on board your excellent comments and hopefully the manuscript is much improved. This is a novel piece of work and there are still many questions and hurdles to overcome but we hope this paper will present the fundamental basis for the approach.

Major comments:

1.)    Some context is needed to explain why the cost distance is used. What’s the advantage and how it works in general?   How are the cost surface raster maps with different variables (current diction, FUI values) used in the cost distance calculation?  

Authors response: Additional detail has been provided with regard to the use of the distance accumulation approach to provide context – see lines lines 198-206. Implementation of the cost surface into the distance accumulation model is provided in section 2.5.2 for example.

 

2.)    While the probability maps of DIN distribution can be derived, no actual validation is carried out in this study.  How do the authors know their results are realistic?

Authors response: Our study represents the first attempt at a probabilistic approach and in situ data is key for validation. We agree that validation of the model is an issue and we have indicated this in the discussion – see lines 479-496 and Conclusions 555-573. To have the same temporal and spatial reach, the obvious candidate for validation would be modelled outputs. However, these also suffer with a lack of in situ measurements and have issues. This is an area we will explore in the future.

 

3.)    How the current velocity/direction affect your results?  Current direction is used to construct cost surface. Why the current speed is not used? How accurate are the reanalysis ocean current datasets?  A discussion on this should be given.  

Authors response: Current speeds were not used because of concerns over the accuracy of them in the near shore waters. The accuracy of the modelled currents (direction) is covered in the limitations section in the discussion – see lines 479-496. We plan to look at trying to improve this in future iterations of the model.

 

4.)    Regarding the forces governing the distribution of DIN, what are the roles of the winds, tidal induced mixing, and fresh water transport in this study? The effects of those are not included in the algorithm (at least not in constructing the cost surface).  

Authors response: Because the DIN cost surface employs a quasi- in situ measurement of the DIN concentration at each given location the influence of tidal patterns and wind are already incorporated into the spatial distribution observed. Furthermore, note that the residual currents model used is forced by tidal, wind and freshwater influences and therefore by using the current direction are considered within the process. For the riverine input cost distance this at present only incorporates the coarse current direction into the modelling. Adapting the process to include tidal and wind mixing would not be a trivial process. For example, to incorporate a tidal cycle would require partitioning of the data into tidal windows and for each tidal window having a daily cost distance raster generated. With the satellite data not able to guarantee daily FUI data this does not seem to be a feasible approach.

 

Minor Comments:

1.)    Check the figure longitude/latitude labels to make sure they are consistent with different figures and clear visible.  

Authors response: These have been checked and the grid for Figure 8a and Figure 8b made consistent with the other figures.

 

2.)    Line 541: “normalized cost surfaces can be found in the GitHub repository”, but the git link shows only a README file.

Authors response: Apologies – the link was to a new repository which has been updated with the code and html markdown output.

3.)    Some acronyms need to be spell out, such EO at line 124, OSPAR at line 90 etc.

Authors response: These acronyms have been spelt out on first position in manuscript.

4.)    Line 178, the current is interpolated from model data with nearest neighbor approach. Please give the original reanalysis/model resolution. If the model resolution is coarse, the nearest neighbor method may have issue.   

Authors response:  The details of the original rasters (from netCDF files) has been added and the processing checked (see below) and the resampling was bilinear and not NN – this has been amended lines 183-185.

# This script generates the direction rasters by resampling

import os

import arcpy

from arcpy import env

from arcpy.sa import *

 

# These are the default locations and names

MonthList = ["Jan","Feb","Mar","Apr","May","Jun","Jul","Aug","Sep","Oct","Nov","Dec"]

MonthList2 = ["01","02","03","04","05","06","07","08","09","10","11","12"]

SeasonList = ["Spring", "Summer", "Autumn", "Winter"]

DefaultDB = r'C:\Users\XXX\Documents\ArcGIS\Projects\RiskMapping_NewData_Paper\RiskMapping_NewData_Paper.gdb'

PDB = r'C:\Users\XXX\Documents\ArcGIS\Projects\RiskMapping_NewData_Paper\PlumeBoundaryGeneration.gdb'

DirectionRasters = r'C:\Users\XXX\Documents\ArcGIS\Projects\RiskMapping_NewData_Paper\Rasters\DirectionRasters'

DirRasterGDB = r'C:\Users\XXX\Documents\ArcGIS\Projects\RiskMapping_NewData_Paper\DirectionRasters.gdb'

 

# Set environment details

# Set the snap raster to ensure alignment

arcpy.env.snapRaster = os.path.join(DefaultDB, "PlumeBoundaryRaster_20172021")

arcpy.env.addOutputsToMap = False

 

# Resampling of rasters to the same resolution as the FUI rasters

for m in MonthList:

             print(["Processing direction raster for month: " + m])   

             # Get the direction raster for plume 1 to resample

             dir_rast = os.path.join(DirectionRasters, ("Plume1_dir_" + m + ".tif"))

             # Sort out the raster to save to

             out_rast = os.path.join(DirRasterGDB, ("Plume1_dir_" + m))

             # Resample the raster

             arcpy.management.Resample(dir_rast, out_rast, "0.003 0.003", "BILINEAR")

             # Get the direction raster for plume 1 to resample

             dir_rast = os.path.join(DirectionRasters, ("Plume2_dir_" + m + ".tif"))

             # Sort out the raster to save to

             out_rast = os.path.join(DirRasterGDB, ("Plume2_dir_" + m))

             # Resample the raster

             arcpy.management.Resample(dir_rast, out_rast, "0.003 0.003", "BILINEAR")

 

print("Complete")

5.)    Conclusion section needs rewritten; it is more like still in discussion.

Authors response: The conclusion section has been rewritten and is hopefully more agreeable.

Round 2

Reviewer 1 Report

Authors attended all my comments and suggestions. Missed references and methodological details have been completed. Limitations of the study has been explained in the text. 

Author Response

Thank you for reading the manuscript again and considering our responses. We are pleased that you are content with the responses.  Many thanks for your time helping to improve the manuscript.

Reviewer 2 Report

There is a point that needs to be corrected in the article. According to comment 5 in the first round, some main and important papers from literature about "water quality" are suggested. The authors replied this comment only by “Thank you for directing us to these papers although they are over 10 years old …”. The suggested papers are not added in the paper and this part is still missing. All suggested papers about water quality should be mentioned and discussed in the paper.

Author Response

Thank you for reading our manuscript again and considering our responses. As requested, we have included the water quality references into the manuscript as appropriate – references 16 & 21; see lines 52 & 431. We trust that this has improved the manuscript to your satisfaction, and we thank you for helping to improve the manuscript.

Reviewer 3 Report

The revised paper addressed my concerns regarding the first version.

Author Response

Thank you for reading the manuscript again and considering our responses. We are pleased that you are content with the responses, and we would like to thank you for your time helping to improve the manuscript.