the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 09 Jan 2025
Phytoplankton detection study through hyperspectral signals in Patagonian Fjords
Abstract. Over recent decades, monitoring coastal areas has becoming a priority due to human population pressure. These areas constitute biodiversity enclaves where the increment in phytoplankton blooms have become a socio-ecological problem with severe impacts at global and regional scales. An important area they affect is the Patagonia Fjords, a complex and intricate coastal system strongly exposed to climate forcings and anthropogenic impacts with the aquaculture industry (salmon and mussel farming) as the main source of income. Within this context, fast and accurate monitoring of phytoplankton in the area is crucial. In this study, we focus on using a new technology combining hyperspectral sensors and unmanned aerial vehicles (UAVs) to detect, identify, and differentiate phytoplankton species from optical data. Findings show differences not only between diatoms and dinoflagellates through the shape and magnitude of the spectral signal at 440, 470, 500, 520, 550, 570, and 580 nm but also at the genera level (Rhizosolenia sp., Pseudo-nitzschia sp., Skeletonema sp., Chaetoceros sp., and Leptocylindrus sp.) and even species level like Heterocapsa triquetra. Chlorophyll-a concentration played a key role in reflectance spectra showing high variability in the green-red bands (~ 500–750 nm) at low concentrations (< 2 μg L-1), and even greater at the blue bands (~ 400–490 nm) under higher concentrations (> 4 μg L-1). Although this work presents a step forward in using new tools and monitoring methodology of phytoplankton in complex coastal systems with a new identification route, more high-quality data from a wide range of ecosystems and environments is still necessary.
-
Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
-
Preprint
(2728 KB)
-
Supplement
(115 KB)
-
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2728 KB) - Metadata XML
-
Supplement
(115 KB) - BibTeX
- EndNote
- Final revised paper
Journal article(s) based on this preprint
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-3951', Anonymous Referee #1, 24 Feb 2025
-
CC1: 'Reply on RC1', Pilar Aparicio, 06 Mar 2025
To whom it may concern,
Thank you very much for your contributions to the improvement of the manuscript “Phytoplankton detection study through hyperspectral signals in Patagonian Fjords” We would like to inform you that all your general and specific comments have been considered.
Concerning the main comments:
1.- Improvement in the introduction regarding satellite limitations and hyperspectral advantages: The text included in lines 58-67 and 100-109 seeks to enhance the information in the introduction but expands the discussion in the corresponding section (discussion).
2.- Regarding data processing (the noise reduction process, geometric correction process): A more detailed text has been added (lines 186-193). However, we’d like to mention that in our study, only reflectance data was used, not hyperspectral images.
3.- Respect to a comparison and/or validation process matching in situ reflectance data with satellite. This aspect has been considered for us, but we had several problems with the scarcity of data available for our monitored coastal area on the punctual day. We are working to obtain satellite data from different missions with a greater spatial and temporal resolution. Regarding comparison or validation with in situ data, please consider that all data we used in this study have been taken in situ.
4.- Thanks to this comment, all the figures and tables of the paper have been revised. Changes suggested by Figure 1 have been applied, and the figure caption has also been corrected. Tables decimals have been unified. Figures 3 and 4 colours have been corrected, and Figure 5 Chlorophyll acronym has been corrected.
5.- In the discussion section, a more detailed comparison with other studies around the spectral signal has been introduced (lines 357-369).
In relation to the specific comments, all of them have been considered in general. Only to mention:
- Line 145. The sentence implies that Chl-a was measured in the microplankton range, which is likely not the case. Please revise.
In this case, we corrected the text (line 162) to be clear that phytoplankton species composition analysis has been done under the microphytoplankton range.
- Lines 382-385. These two sentences lack logical flow. The first sentence discusses non-biological particles, while the second shifts to phytoplankton.
We have changed both text and paragraph order (lines 405-415), trying to improve the flow of the writing.
We want to mention that all cites along the text, mainly in the introduction and discussion sections, have been revised and adapted to reduce the number and updated when it corresponds. Finally, we would like to mention that all the changes applied to the manuscript have been highlighted in red and yellow to make it easier and faster to review them.
Thank you very much for your time and contributions.
Sincerely, Dra. P.A.
-
AC1: 'Reply on RC2', Daniel Varela, 01 Apr 2025
Dear reviewer ,
Thank you very much for your constructive comments to manuscript. We are trying to consider all your comments.
Concerning your main comments:
1.- First, I found some parts in the methodology was not sufficiently detailed. For example, the 2.5 Data processing (should be named as Data Analyses at least) section has very limited information, inconsistent with the level of details in other subsections 2.1-2.4. It is unclear if the authors have conducted geometric corrections, radiometric corrections (and how), HCA and PCA analysis steps, etc. When I read Section 3 Results, I found it difficult to match with Section 2. Also, where is the results for the non-parametric K-S test?
- Regarding to the subtitle, we consider thar both meaning are relevant, because of that we decide use “Data processing and analyses”.
- Regarding data processing (the noise reduction process, and geometric correction process): A more detailed text has been added. the paragraph in this section would read as follows:
“Several steps were involved in processing the raw image data captured by the hyperspectral camera using the Spectronon Pro software. First, the raw data were processed to obtain reflectance (R(λ)), using the calibration file provided by Resonon and the downwelling irradiance data provided by Miniature Spectrometer (Flame Ocean Insight). In the following step, pixel cleaning by an adapted NDWI index was undertaken, masking those pixels that did not correspond to water and those saturated due to solar radiation angle incidence (Xie et al., 2014). The lack of reference points in the monitoring areas a no geometric correction was applied to hyperspectral data, because of that the data are used as data instead imagen in this study. In the next step, this data was normalize using min-max method, such that the new minimum and maximum values are 0 and 1. After that, the use of Savitzky-Golay method or digital smoothing polynomial filtering (DISPO), based on least squares polynomial smoothing fit and differentiation (Ruffin and King, 1999; Gallagher, 2020), was applied to reduce the signal of the noise and obtain a clean spectrum, with less distortion than moving average techniques, without altering its properties. Finally, reflectance raw data were averaged, obtaining a reflectance value at each band to emphasise the shape singularities of each spectrum.”
- The non-parametric K-S test data have been added at a supplementary table referred as Table S2 in the results texts (lines 276, 283-287, 290-2).
2.- Second, my main critique is about Section 3.1, which characterizes those sites but takes a few pages. While it is helpful to characterize those sites, it was done with lengthy statistical analyses and their visualizations that are not quite relevant to the main theme - hyperspectral analyses. It is unclear how it supports the main objective "the characterisation of the reflectance spectra of different phytoplankton assemblages, either harmful or non-harmful, dominated by a single species." However, if this subsection were removed, the results section would be so slim.
- Although the description of environmental condition is a bit extensive and somewhat irrelevant for the main objective, we consider an important aspect in the text, especially to determine if these conditions indirectly affect the spectral signal. Indeed, an apparent relation between the spectral signals and localities condition, especially regarding to the chlorophyl.
3.- Third, because the authors focused on a pilot study in this article, it is reasonable to expect a comprehensive blueprint on the path forward. I found Section 4.2 was relatively underdeveloped. Challenges and opportunities are better to be organized around a few general themes for deeper insights.
- Regarding to this point, you can see in the discussion 4.2 section, changes at text expressing the principal challenges and advantages that our equipment have experienced during the monitoring. Thus, you will read in the manuscript:
“Other limitations that appeared frequently during the study were related to technical aspects such as the flight time limited by the battery’s life, the area coverage limited by network signal and again batteries durability, or the lack of reference points in a continuous body water (Kislik et al., 2018; Wu et al., 2019). The battery’s life was identified as a quite important aspect being the main factor that spatially limited the monitoring to coastal areas and also few stations barely separated during the study. Another aspect to consider is the camera size. Nowadays hyperspectral cameras are still new being not yet completely functional due to their big size and weight that force to acquire a bigger drone and therefore a more expensive one. Related to the camera one more aspect that can be a challenge is the high variability at cameras market, with different characteristics in spectral resolution, geometry acquisition and imaging systems. Although this can be considered at first as a positive point, we have to be cautious at the moment to acquire the camera because this is going to limit the possibility to compare data and apply algorithms in future.
Other important barriers which make it difficult to identify phytoplankton using reflectance data are the analysis of the vertical distribution of phytoplankton in the water column (~ 1-40 m) in both satellites and UAV systems, especially in the design of future algorithms. Therefore, not just the phytoplankton distribution in water column but also the presence of other particles, mainly non-biological ones, and their impact in solar radiation absorption/reflexion together with physical aspects as the water roughness.
Although recent works have been including hyperspectral data for the study of phytoplankton assemblages, these datasets are still limited to specific areas and unavailable, being the lack of a robust hyperspectral images’ library another important factor to consider.”
Regarding your minor comments
Lastly, the article suffers from the use of inaccurate phrases, grammatical errors, and labeling issues.
- For example, at Line55, spatial-temporal resolution has contained the long revisit cycle. à
- This text has been rephase and grammatical reviewed by a native British.
- At Line 82, it is unclear if geography (complex topography) would be a direct issue for phytoplankton.
- Although the reference to the complex topography (geography) was not related to phytoplankton, we have had changes to be clearer in the text.
- In the legend of Fig 11, the label for low and high Chlo(s) were not distinguishable.
- We have reviewed and corrected the Chl-a label at both in the legend of the figure 11 and inside the figure. Now the Chl-a labels are higher.
- Line 344: the phrase 'a non-significant difference' sounds odd.
- This has been reviewed and corrected.
- Line 350: what is "clusters together assemblage"?
- This expression refers that the HCA groups localities under same phytoplankton groups and genera dominance. The statement was reviewed and improved.
- Line 366: variability -> the variability.
- Reviewed and corrected.
Thank you very much for your time and contributions.
Citation: https://doi.org/10.5194/egusphere-2024-3951-AC1
-
CC1: 'Reply on RC1', Pilar Aparicio, 06 Mar 2025
-
RC2: 'Comment on egusphere-2024-3951', Anonymous Referee #2, 13 Mar 2025
This article discusses a pilot study on using hyperspectral drone imagery to monitor different species of phytoplankton. It is an interesting effort as distinguishing diverse species outside a lab environment is always a challenge. This paper mainly uses preliminary results as an entry point to discuss challenges and opportunities of this direction. I have a few major and minor comments that I hope would be constructive to the authors.
First, I found some parts in the methodology was not sufficiently detailed. For example, the 2.5 Data processing (should be named as Data Analyses at least) section has very limited information, inconsistent with the level of details in other subsections 2.1-2.4. It is unclear if the authors have conducted geometric corrections, radiometric corrections (and how), HCA and PCA analysis steps, etc. When I read Section 3 Results, I found it difficult to match with Section 2. Also, where is the results for the non-parametric K-S test?
Second, my main critique is about Section 3.1, which characterizes those sites but takes a few pages. While it is helpful to characterize those sites, it was done with lengthy statistical analyses and their visualizations that are not quite relevant to the main theme - hyperspectral analyses. It is unclear how it supports the main objective "the characterisation of the reflectance spectra of different phytoplankton assemblages, either harmful or non-harmful, dominated by a single species." However, if this subsection were removed, the results section would be so slim.
Third, because the authors focused on a pilot study in this article, it is reasonable to expect a comprehensive blueprint on the path forward. I found Section 4.2 was relatively underdeveloped. Challenges and opportunities are better to be organized around a few general themes for deeper insights.
Lastly, the article suffers from the use of inaccurate phrases, grammatical errors, and labeling issues. For example, at Line55, spatio-temporal resolution has contained the long revisit cycle. At Line 82, it is unclear if geography (complex topography) would be a direct issue for phytoplankton. In the legend of Fig 11, the label for low and high Chlo(s) were not distinguishable. Line 344: the phrase 'a non-significant difference' sounds odd. Line 350: what is "clusters together assemblage"? Line 366: variability -> the variability.
Citation: https://doi.org/10.5194/egusphere-2024-3951-RC2 -
AC1: 'Reply on RC2', Daniel Varela, 01 Apr 2025
Dear reviewer ,
Thank you very much for your constructive comments to manuscript. We are trying to consider all your comments.
Concerning your main comments:
1.- First, I found some parts in the methodology was not sufficiently detailed. For example, the 2.5 Data processing (should be named as Data Analyses at least) section has very limited information, inconsistent with the level of details in other subsections 2.1-2.4. It is unclear if the authors have conducted geometric corrections, radiometric corrections (and how), HCA and PCA analysis steps, etc. When I read Section 3 Results, I found it difficult to match with Section 2. Also, where is the results for the non-parametric K-S test?
- Regarding to the subtitle, we consider thar both meaning are relevant, because of that we decide use “Data processing and analyses”.
- Regarding data processing (the noise reduction process, and geometric correction process): A more detailed text has been added. the paragraph in this section would read as follows:
“Several steps were involved in processing the raw image data captured by the hyperspectral camera using the Spectronon Pro software. First, the raw data were processed to obtain reflectance (R(λ)), using the calibration file provided by Resonon and the downwelling irradiance data provided by Miniature Spectrometer (Flame Ocean Insight). In the following step, pixel cleaning by an adapted NDWI index was undertaken, masking those pixels that did not correspond to water and those saturated due to solar radiation angle incidence (Xie et al., 2014). The lack of reference points in the monitoring areas a no geometric correction was applied to hyperspectral data, because of that the data are used as data instead imagen in this study. In the next step, this data was normalize using min-max method, such that the new minimum and maximum values are 0 and 1. After that, the use of Savitzky-Golay method or digital smoothing polynomial filtering (DISPO), based on least squares polynomial smoothing fit and differentiation (Ruffin and King, 1999; Gallagher, 2020), was applied to reduce the signal of the noise and obtain a clean spectrum, with less distortion than moving average techniques, without altering its properties. Finally, reflectance raw data were averaged, obtaining a reflectance value at each band to emphasise the shape singularities of each spectrum.”
- The non-parametric K-S test data have been added at a supplementary table referred as Table S2 in the results texts (lines 276, 283-287, 290-2).
2.- Second, my main critique is about Section 3.1, which characterizes those sites but takes a few pages. While it is helpful to characterize those sites, it was done with lengthy statistical analyses and their visualizations that are not quite relevant to the main theme - hyperspectral analyses. It is unclear how it supports the main objective "the characterisation of the reflectance spectra of different phytoplankton assemblages, either harmful or non-harmful, dominated by a single species." However, if this subsection were removed, the results section would be so slim.
- Although the description of environmental condition is a bit extensive and somewhat irrelevant for the main objective, we consider an important aspect in the text, especially to determine if these conditions indirectly affect the spectral signal. Indeed, an apparent relation between the spectral signals and localities condition, especially regarding to the chlorophyl.
3.- Third, because the authors focused on a pilot study in this article, it is reasonable to expect a comprehensive blueprint on the path forward. I found Section 4.2 was relatively underdeveloped. Challenges and opportunities are better to be organized around a few general themes for deeper insights.
- Regarding to this point, you can see in the discussion 4.2 section, changes at text expressing the principal challenges and advantages that our equipment have experienced during the monitoring. Thus, you will read in the manuscript:
“Other limitations that appeared frequently during the study were related to technical aspects such as the flight time limited by the battery’s life, the area coverage limited by network signal and again batteries durability, or the lack of reference points in a continuous body water (Kislik et al., 2018; Wu et al., 2019). The battery’s life was identified as a quite important aspect being the main factor that spatially limited the monitoring to coastal areas and also few stations barely separated during the study. Another aspect to consider is the camera size. Nowadays hyperspectral cameras are still new being not yet completely functional due to their big size and weight that force to acquire a bigger drone and therefore a more expensive one. Related to the camera one more aspect that can be a challenge is the high variability at cameras market, with different characteristics in spectral resolution, geometry acquisition and imaging systems. Although this can be considered at first as a positive point, we have to be cautious at the moment to acquire the camera because this is going to limit the possibility to compare data and apply algorithms in future.
Other important barriers which make it difficult to identify phytoplankton using reflectance data are the analysis of the vertical distribution of phytoplankton in the water column (~ 1-40 m) in both satellites and UAV systems, especially in the design of future algorithms. Therefore, not just the phytoplankton distribution in water column but also the presence of other particles, mainly non-biological ones, and their impact in solar radiation absorption/reflexion together with physical aspects as the water roughness.
Although recent works have been including hyperspectral data for the study of phytoplankton assemblages, these datasets are still limited to specific areas and unavailable, being the lack of a robust hyperspectral images’ library another important factor to consider.”
Regarding your minor comments
Lastly, the article suffers from the use of inaccurate phrases, grammatical errors, and labeling issues.
- For example, at Line55, spatial-temporal resolution has contained the long revisit cycle. à
- This text has been rephase and grammatical reviewed by a native British.
- At Line 82, it is unclear if geography (complex topography) would be a direct issue for phytoplankton.
- Although the reference to the complex topography (geography) was not related to phytoplankton, we have had changes to be clearer in the text.
- In the legend of Fig 11, the label for low and high Chlo(s) were not distinguishable.
- We have reviewed and corrected the Chl-a label at both in the legend of the figure 11 and inside the figure. Now the Chl-a labels are higher.
- Line 344: the phrase 'a non-significant difference' sounds odd.
- This has been reviewed and corrected.
- Line 350: what is "clusters together assemblage"?
- This expression refers that the HCA groups localities under same phytoplankton groups and genera dominance. The statement was reviewed and improved.
- Line 366: variability -> the variability.
- Reviewed and corrected.
Thank you very much for your time and contributions.
Citation: https://doi.org/10.5194/egusphere-2024-3951-AC1
-
AC1: 'Reply on RC2', Daniel Varela, 01 Apr 2025
Interactive discussion
Status: closed
-
RC1: 'Comment on egusphere-2024-3951', Anonymous Referee #1, 24 Feb 2025
Review of “Phytoplankton detection study through hyperspectral signals in Patagonian Fjords”
The manuscript by Aparicio-Rizzo et al. presents an interesting study on the use of hyperspectral sensors mounted on UAVs to monitor phytoplankton in the complex coastal systems of the Patagonian Fjords. The manuscript has a clear objective, that is to address the challenges in monitoring phytoplankton blooms using in-situ or satellite monitoring methods in complex coastal areas, particularly in regions with harmful algal blooms. Based on the collected UAV dataset, the authors have demonstrated the potential of hyperspectral data to differentiate between phytoplankton species and genera, which is an advancement in the field of remote sensing and phytoplankton monitoring.
Overall, the manuscript is good-structured and clearly written. The hyperspectral data and its concurrent oceanographic measurements provide a comprehensive understanding of the phytoplankton dynamics in the study area. Moreover, the use of UAVs for in-situ reflectance measurements is innovative, offering a practical solution to the limitations of satellite-based monitoring in regions with high cloud cover and complex coastal environments.
However, there are a few areas where the manuscript could be improved. Below, I provide general and specific comments aimed at improving the manuscript.
General comments
- The introduction provides a comprehensive background on the importance of phytoplankton monitoring in the region. However, it would benefit from a more detailed discussion of the limitations of satellite-based monitoring.Additionally, a clearer explanation of the advantages of hyperspectral imaging over traditional methods would also help.
- The manuscript mentions the calibration of the hyperspectral sensor and the calculation of remote sensing reflectance. However, it lacks detailed steps on how these processes were carried out. For example:
- detailed steps or results of the noise reduction process;
- more details on the geometric correction process;
- a detailed assessment of data quality, e.g., noise levels before and after processing.
- Do you have any UAV-matched satellite data or in-situ measurements? Including a comparison or validation with matched satellite or in-situ data would strengthen the findings.
- The criteria for selecting sampling locations are well explained, but the presentation of these locations in the figures could be improved. For example, in Figure 1, it is unclear how (b) and (c) relate to (a), and the labelling of stations (e.g., TENCLO) does not clearly match across figures (e.g., Figures 3-5) and Table 1. It is better to ensure consistent labelling, which would help readers unfamiliar with the region better understand it.
- The discussion could benefit from a more detailed comparison with existing studies. For example, there are many studies on the spectral properties of dinoflagellates and diatoms (through in-situ or satellite observation). Additionally, some paragraphs rely heavily on the same references, please consider selecting the most appropriate references.
Specific comments
- Lines 38-58. As mentioned above, the references used in this section are not always appropriate. Consider revising to include more relevant studies or selecting the most appropriate references to support the points being made.
- Line 70. CDOM is not particulate matter.This should be revised.
- Line 140. The sentence could be revised for clarity.
- Line 145. The sentence implies that Chl-a was measured in the microplankton range, which is likely not the case. Please revise.
- Line 173. Add a comma after "signal".
- Table 2. Ensure consistency in decimal places across temperature, salinity, and Chl-a values.
- Figure 5. Recommend changing "Chlo" to "Chl-a" to be consistent with the rest of the manuscript.
- Lines 382-385. These two sentences lack logical flow. The first sentence discusses non-biological particles, while the second shifts to phytoplankton.
Citation: https://doi.org/10.5194/egusphere-2024-3951-RC1 -
CC1: 'Reply on RC1', Pilar Aparicio, 06 Mar 2025
To whom it may concern,
Thank you very much for your contributions to the improvement of the manuscript “Phytoplankton detection study through hyperspectral signals in Patagonian Fjords” We would like to inform you that all your general and specific comments have been considered.
Concerning the main comments:
1.- Improvement in the introduction regarding satellite limitations and hyperspectral advantages: The text included in lines 58-67 and 100-109 seeks to enhance the information in the introduction but expands the discussion in the corresponding section (discussion).
2.- Regarding data processing (the noise reduction process, geometric correction process): A more detailed text has been added (lines 186-193). However, we’d like to mention that in our study, only reflectance data was used, not hyperspectral images.
3.- Respect to a comparison and/or validation process matching in situ reflectance data with satellite. This aspect has been considered for us, but we had several problems with the scarcity of data available for our monitored coastal area on the punctual day. We are working to obtain satellite data from different missions with a greater spatial and temporal resolution. Regarding comparison or validation with in situ data, please consider that all data we used in this study have been taken in situ.
4.- Thanks to this comment, all the figures and tables of the paper have been revised. Changes suggested by Figure 1 have been applied, and the figure caption has also been corrected. Tables decimals have been unified. Figures 3 and 4 colours have been corrected, and Figure 5 Chlorophyll acronym has been corrected.
5.- In the discussion section, a more detailed comparison with other studies around the spectral signal has been introduced (lines 357-369).
In relation to the specific comments, all of them have been considered in general. Only to mention:
- Line 145. The sentence implies that Chl-a was measured in the microplankton range, which is likely not the case. Please revise.
In this case, we corrected the text (line 162) to be clear that phytoplankton species composition analysis has been done under the microphytoplankton range.
- Lines 382-385. These two sentences lack logical flow. The first sentence discusses non-biological particles, while the second shifts to phytoplankton.
We have changed both text and paragraph order (lines 405-415), trying to improve the flow of the writing.
We want to mention that all cites along the text, mainly in the introduction and discussion sections, have been revised and adapted to reduce the number and updated when it corresponds. Finally, we would like to mention that all the changes applied to the manuscript have been highlighted in red and yellow to make it easier and faster to review them.
Thank you very much for your time and contributions.
Sincerely, Dra. P.A.
-
AC1: 'Reply on RC2', Daniel Varela, 01 Apr 2025
Dear reviewer ,
Thank you very much for your constructive comments to manuscript. We are trying to consider all your comments.
Concerning your main comments:
1.- First, I found some parts in the methodology was not sufficiently detailed. For example, the 2.5 Data processing (should be named as Data Analyses at least) section has very limited information, inconsistent with the level of details in other subsections 2.1-2.4. It is unclear if the authors have conducted geometric corrections, radiometric corrections (and how), HCA and PCA analysis steps, etc. When I read Section 3 Results, I found it difficult to match with Section 2. Also, where is the results for the non-parametric K-S test?
- Regarding to the subtitle, we consider thar both meaning are relevant, because of that we decide use “Data processing and analyses”.
- Regarding data processing (the noise reduction process, and geometric correction process): A more detailed text has been added. the paragraph in this section would read as follows:
“Several steps were involved in processing the raw image data captured by the hyperspectral camera using the Spectronon Pro software. First, the raw data were processed to obtain reflectance (R(λ)), using the calibration file provided by Resonon and the downwelling irradiance data provided by Miniature Spectrometer (Flame Ocean Insight). In the following step, pixel cleaning by an adapted NDWI index was undertaken, masking those pixels that did not correspond to water and those saturated due to solar radiation angle incidence (Xie et al., 2014). The lack of reference points in the monitoring areas a no geometric correction was applied to hyperspectral data, because of that the data are used as data instead imagen in this study. In the next step, this data was normalize using min-max method, such that the new minimum and maximum values are 0 and 1. After that, the use of Savitzky-Golay method or digital smoothing polynomial filtering (DISPO), based on least squares polynomial smoothing fit and differentiation (Ruffin and King, 1999; Gallagher, 2020), was applied to reduce the signal of the noise and obtain a clean spectrum, with less distortion than moving average techniques, without altering its properties. Finally, reflectance raw data were averaged, obtaining a reflectance value at each band to emphasise the shape singularities of each spectrum.”
- The non-parametric K-S test data have been added at a supplementary table referred as Table S2 in the results texts (lines 276, 283-287, 290-2).
2.- Second, my main critique is about Section 3.1, which characterizes those sites but takes a few pages. While it is helpful to characterize those sites, it was done with lengthy statistical analyses and their visualizations that are not quite relevant to the main theme - hyperspectral analyses. It is unclear how it supports the main objective "the characterisation of the reflectance spectra of different phytoplankton assemblages, either harmful or non-harmful, dominated by a single species." However, if this subsection were removed, the results section would be so slim.
- Although the description of environmental condition is a bit extensive and somewhat irrelevant for the main objective, we consider an important aspect in the text, especially to determine if these conditions indirectly affect the spectral signal. Indeed, an apparent relation between the spectral signals and localities condition, especially regarding to the chlorophyl.
3.- Third, because the authors focused on a pilot study in this article, it is reasonable to expect a comprehensive blueprint on the path forward. I found Section 4.2 was relatively underdeveloped. Challenges and opportunities are better to be organized around a few general themes for deeper insights.
- Regarding to this point, you can see in the discussion 4.2 section, changes at text expressing the principal challenges and advantages that our equipment have experienced during the monitoring. Thus, you will read in the manuscript:
“Other limitations that appeared frequently during the study were related to technical aspects such as the flight time limited by the battery’s life, the area coverage limited by network signal and again batteries durability, or the lack of reference points in a continuous body water (Kislik et al., 2018; Wu et al., 2019). The battery’s life was identified as a quite important aspect being the main factor that spatially limited the monitoring to coastal areas and also few stations barely separated during the study. Another aspect to consider is the camera size. Nowadays hyperspectral cameras are still new being not yet completely functional due to their big size and weight that force to acquire a bigger drone and therefore a more expensive one. Related to the camera one more aspect that can be a challenge is the high variability at cameras market, with different characteristics in spectral resolution, geometry acquisition and imaging systems. Although this can be considered at first as a positive point, we have to be cautious at the moment to acquire the camera because this is going to limit the possibility to compare data and apply algorithms in future.
Other important barriers which make it difficult to identify phytoplankton using reflectance data are the analysis of the vertical distribution of phytoplankton in the water column (~ 1-40 m) in both satellites and UAV systems, especially in the design of future algorithms. Therefore, not just the phytoplankton distribution in water column but also the presence of other particles, mainly non-biological ones, and their impact in solar radiation absorption/reflexion together with physical aspects as the water roughness.
Although recent works have been including hyperspectral data for the study of phytoplankton assemblages, these datasets are still limited to specific areas and unavailable, being the lack of a robust hyperspectral images’ library another important factor to consider.”
Regarding your minor comments
Lastly, the article suffers from the use of inaccurate phrases, grammatical errors, and labeling issues.
- For example, at Line55, spatial-temporal resolution has contained the long revisit cycle. à
- This text has been rephase and grammatical reviewed by a native British.
- At Line 82, it is unclear if geography (complex topography) would be a direct issue for phytoplankton.
- Although the reference to the complex topography (geography) was not related to phytoplankton, we have had changes to be clearer in the text.
- In the legend of Fig 11, the label for low and high Chlo(s) were not distinguishable.
- We have reviewed and corrected the Chl-a label at both in the legend of the figure 11 and inside the figure. Now the Chl-a labels are higher.
- Line 344: the phrase 'a non-significant difference' sounds odd.
- This has been reviewed and corrected.
- Line 350: what is "clusters together assemblage"?
- This expression refers that the HCA groups localities under same phytoplankton groups and genera dominance. The statement was reviewed and improved.
- Line 366: variability -> the variability.
- Reviewed and corrected.
Thank you very much for your time and contributions.
Citation: https://doi.org/10.5194/egusphere-2024-3951-AC1
-
RC2: 'Comment on egusphere-2024-3951', Anonymous Referee #2, 13 Mar 2025
This article discusses a pilot study on using hyperspectral drone imagery to monitor different species of phytoplankton. It is an interesting effort as distinguishing diverse species outside a lab environment is always a challenge. This paper mainly uses preliminary results as an entry point to discuss challenges and opportunities of this direction. I have a few major and minor comments that I hope would be constructive to the authors.
First, I found some parts in the methodology was not sufficiently detailed. For example, the 2.5 Data processing (should be named as Data Analyses at least) section has very limited information, inconsistent with the level of details in other subsections 2.1-2.4. It is unclear if the authors have conducted geometric corrections, radiometric corrections (and how), HCA and PCA analysis steps, etc. When I read Section 3 Results, I found it difficult to match with Section 2. Also, where is the results for the non-parametric K-S test?
Second, my main critique is about Section 3.1, which characterizes those sites but takes a few pages. While it is helpful to characterize those sites, it was done with lengthy statistical analyses and their visualizations that are not quite relevant to the main theme - hyperspectral analyses. It is unclear how it supports the main objective "the characterisation of the reflectance spectra of different phytoplankton assemblages, either harmful or non-harmful, dominated by a single species." However, if this subsection were removed, the results section would be so slim.
Third, because the authors focused on a pilot study in this article, it is reasonable to expect a comprehensive blueprint on the path forward. I found Section 4.2 was relatively underdeveloped. Challenges and opportunities are better to be organized around a few general themes for deeper insights.
Lastly, the article suffers from the use of inaccurate phrases, grammatical errors, and labeling issues. For example, at Line55, spatio-temporal resolution has contained the long revisit cycle. At Line 82, it is unclear if geography (complex topography) would be a direct issue for phytoplankton. In the legend of Fig 11, the label for low and high Chlo(s) were not distinguishable. Line 344: the phrase 'a non-significant difference' sounds odd. Line 350: what is "clusters together assemblage"? Line 366: variability -> the variability.
Citation: https://doi.org/10.5194/egusphere-2024-3951-RC2 -
AC1: 'Reply on RC2', Daniel Varela, 01 Apr 2025
Dear reviewer ,
Thank you very much for your constructive comments to manuscript. We are trying to consider all your comments.
Concerning your main comments:
1.- First, I found some parts in the methodology was not sufficiently detailed. For example, the 2.5 Data processing (should be named as Data Analyses at least) section has very limited information, inconsistent with the level of details in other subsections 2.1-2.4. It is unclear if the authors have conducted geometric corrections, radiometric corrections (and how), HCA and PCA analysis steps, etc. When I read Section 3 Results, I found it difficult to match with Section 2. Also, where is the results for the non-parametric K-S test?
- Regarding to the subtitle, we consider thar both meaning are relevant, because of that we decide use “Data processing and analyses”.
- Regarding data processing (the noise reduction process, and geometric correction process): A more detailed text has been added. the paragraph in this section would read as follows:
“Several steps were involved in processing the raw image data captured by the hyperspectral camera using the Spectronon Pro software. First, the raw data were processed to obtain reflectance (R(λ)), using the calibration file provided by Resonon and the downwelling irradiance data provided by Miniature Spectrometer (Flame Ocean Insight). In the following step, pixel cleaning by an adapted NDWI index was undertaken, masking those pixels that did not correspond to water and those saturated due to solar radiation angle incidence (Xie et al., 2014). The lack of reference points in the monitoring areas a no geometric correction was applied to hyperspectral data, because of that the data are used as data instead imagen in this study. In the next step, this data was normalize using min-max method, such that the new minimum and maximum values are 0 and 1. After that, the use of Savitzky-Golay method or digital smoothing polynomial filtering (DISPO), based on least squares polynomial smoothing fit and differentiation (Ruffin and King, 1999; Gallagher, 2020), was applied to reduce the signal of the noise and obtain a clean spectrum, with less distortion than moving average techniques, without altering its properties. Finally, reflectance raw data were averaged, obtaining a reflectance value at each band to emphasise the shape singularities of each spectrum.”
- The non-parametric K-S test data have been added at a supplementary table referred as Table S2 in the results texts (lines 276, 283-287, 290-2).
2.- Second, my main critique is about Section 3.1, which characterizes those sites but takes a few pages. While it is helpful to characterize those sites, it was done with lengthy statistical analyses and their visualizations that are not quite relevant to the main theme - hyperspectral analyses. It is unclear how it supports the main objective "the characterisation of the reflectance spectra of different phytoplankton assemblages, either harmful or non-harmful, dominated by a single species." However, if this subsection were removed, the results section would be so slim.
- Although the description of environmental condition is a bit extensive and somewhat irrelevant for the main objective, we consider an important aspect in the text, especially to determine if these conditions indirectly affect the spectral signal. Indeed, an apparent relation between the spectral signals and localities condition, especially regarding to the chlorophyl.
3.- Third, because the authors focused on a pilot study in this article, it is reasonable to expect a comprehensive blueprint on the path forward. I found Section 4.2 was relatively underdeveloped. Challenges and opportunities are better to be organized around a few general themes for deeper insights.
- Regarding to this point, you can see in the discussion 4.2 section, changes at text expressing the principal challenges and advantages that our equipment have experienced during the monitoring. Thus, you will read in the manuscript:
“Other limitations that appeared frequently during the study were related to technical aspects such as the flight time limited by the battery’s life, the area coverage limited by network signal and again batteries durability, or the lack of reference points in a continuous body water (Kislik et al., 2018; Wu et al., 2019). The battery’s life was identified as a quite important aspect being the main factor that spatially limited the monitoring to coastal areas and also few stations barely separated during the study. Another aspect to consider is the camera size. Nowadays hyperspectral cameras are still new being not yet completely functional due to their big size and weight that force to acquire a bigger drone and therefore a more expensive one. Related to the camera one more aspect that can be a challenge is the high variability at cameras market, with different characteristics in spectral resolution, geometry acquisition and imaging systems. Although this can be considered at first as a positive point, we have to be cautious at the moment to acquire the camera because this is going to limit the possibility to compare data and apply algorithms in future.
Other important barriers which make it difficult to identify phytoplankton using reflectance data are the analysis of the vertical distribution of phytoplankton in the water column (~ 1-40 m) in both satellites and UAV systems, especially in the design of future algorithms. Therefore, not just the phytoplankton distribution in water column but also the presence of other particles, mainly non-biological ones, and their impact in solar radiation absorption/reflexion together with physical aspects as the water roughness.
Although recent works have been including hyperspectral data for the study of phytoplankton assemblages, these datasets are still limited to specific areas and unavailable, being the lack of a robust hyperspectral images’ library another important factor to consider.”
Regarding your minor comments
Lastly, the article suffers from the use of inaccurate phrases, grammatical errors, and labeling issues.
- For example, at Line55, spatial-temporal resolution has contained the long revisit cycle. à
- This text has been rephase and grammatical reviewed by a native British.
- At Line 82, it is unclear if geography (complex topography) would be a direct issue for phytoplankton.
- Although the reference to the complex topography (geography) was not related to phytoplankton, we have had changes to be clearer in the text.
- In the legend of Fig 11, the label for low and high Chlo(s) were not distinguishable.
- We have reviewed and corrected the Chl-a label at both in the legend of the figure 11 and inside the figure. Now the Chl-a labels are higher.
- Line 344: the phrase 'a non-significant difference' sounds odd.
- This has been reviewed and corrected.
- Line 350: what is "clusters together assemblage"?
- This expression refers that the HCA groups localities under same phytoplankton groups and genera dominance. The statement was reviewed and improved.
- Line 366: variability -> the variability.
- Reviewed and corrected.
Thank you very much for your time and contributions.
Citation: https://doi.org/10.5194/egusphere-2024-3951-AC1
-
AC1: 'Reply on RC2', Daniel Varela, 01 Apr 2025
Peer review completion
Journal article(s) based on this preprint
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 713 | 129 | 19 | 861 | 29 | 16 | 38 |
- HTML: 713
- PDF: 129
- XML: 19
- Total: 861
- Supplement: 29
- BibTeX: 16
- EndNote: 38
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Pilar Aparicio-Rizzo
Dagoberto Poblete-Cballero
Cristian Vera-Bastidas
Iván Pérez-Santos
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
- Preprint
(2728 KB) - Metadata XML
-
Supplement
(115 KB) - BibTeX
- EndNote
- Final revised paper
Review of “Phytoplankton detection study through hyperspectral signals in Patagonian Fjords”
The manuscript by Aparicio-Rizzo et al. presents an interesting study on the use of hyperspectral sensors mounted on UAVs to monitor phytoplankton in the complex coastal systems of the Patagonian Fjords. The manuscript has a clear objective, that is to address the challenges in monitoring phytoplankton blooms using in-situ or satellite monitoring methods in complex coastal areas, particularly in regions with harmful algal blooms. Based on the collected UAV dataset, the authors have demonstrated the potential of hyperspectral data to differentiate between phytoplankton species and genera, which is an advancement in the field of remote sensing and phytoplankton monitoring.
Overall, the manuscript is good-structured and clearly written. The hyperspectral data and its concurrent oceanographic measurements provide a comprehensive understanding of the phytoplankton dynamics in the study area. Moreover, the use of UAVs for in-situ reflectance measurements is innovative, offering a practical solution to the limitations of satellite-based monitoring in regions with high cloud cover and complex coastal environments.
However, there are a few areas where the manuscript could be improved. Below, I provide general and specific comments aimed at improving the manuscript.
General comments
Specific comments