the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Hyperspectral reflectance of pristine, ocean weathered and biofouled plastics from dry to wet and submerged state
Robin V. F. de Vries
Shungudzemwoyo P. Garaba
Sarah-Jeanne Royer
Abstract. High-quality spectral reference libraries are important for algorithm development and identification of diagnostic optical features of target objects in environmental remote sensing applications. We present additional measurements conducted using hyperspectral sensor technologies in a laboratory and outdoor setting to further extend high-quality data as well as diversity in available open-access spectral reference libraries. These observations involved gathering hyperspectral single-pixel point and multi-pixel optical properties of a diverse set of plastic materials (e.g., ropes, nets, packaging, and personal protective equipment). Measurements of COVID-19 personal protective equipment were conducted to also further expand reference datasets that could be useful in monitoring mismanaged waste related to the pandemic. The sample set consisted of virgin polymers and ocean-weathered and artificially biofouled objects of varying apparent colors, shapes, forms, thicknesses, and opacity. A Spectral Evolution spectroradiometer was used to collect hyperspectral reflectance single pixel point information from 280 – 2500 nm. Imaging was also performed using a Specim IQ hyperspectral camera from 400 – 1000 nm. Sampling underwater was completed in intervals of 0.005 m to 0.215 m within a 0.005 – 0.715 m depth range. All optical measurements are available in open-access for the laboratory experiment through https://doi.org/10.4121/769cc482-b104-4927-a94b-b16f6618c3b3 (de Vries and Garaba, 2023) and outdoors campaign involving the biofouling samples via https://doi.org/10.4121/7c53b72a-be97-478b-9288-ff9c850de64b (de Vries et al., 2023).
- Preprint
(3964 KB) - Metadata XML
- BibTeX
- EndNote
Robin V. F. de Vries et al.
Status: closed
-
CC1: 'Comment on essd-2023-209', Chuanmin Hu, 17 Jun 2023
This is a great contribution towards remote detection of plastics in natural environments. The data are made available to the community, which is particularly important to reference against other measurements and to develop algorithms.
For this reason, I recommend publication, but I would also suggest some revisions to improve the presentation.
My biggest concern is how these data can be applied to remote sensing applications. In other words, which spectra can be used as the endmember spectra to interpret remote sensing imagery? Why?
There are also ambiguities here and there that should be corrected or clarified. Please see below.
Leone et al. (2023) did very similar experiments except with a limited depth range of 0 – 0.08 m. So what’s the motivation to do it again? Is the additional depth range of 0.08 – 0.7 m that important? If so, why 0.7 m? Actually both Figs. 8 & 9 show that after 0.08 m, there is nearly no information in the SWIR wavelengths.
For the same reason, how are the experimental results compared with those of Leone et al. (2023) for the same depth range of 0 – 0.08 m? Are they the same or different? If they are different, why?
Same question applies to Knaeps et al. (2021). It’s great to list previous efforts to establish spectral libraries in Table 1 (line 44). But so? Are these earlier efforts not enough? In what way?
What’s the purpose to use take the measurement in Fig. 2? Just to record a photograph of each type?
Fig. 3 shows Halogen tungsten lamps but Figs. 5&6 shows natural illumination. Then, for the reported spectral library, which light source was used? Fig. 3 is not cited but described in the text (line 108 – 117). But what’s the purpose of this experimental setting? My experience with lamps is that they don’t provide collimated beams, meaning that the amount of light received by the target (irradiance) depends on the distance between the target and the light source. Then it will introduce errors when calculating reflectance because the target and the reference plaque may be placed in different positions relative to the light source.
After all, which experimental setting was used to measure reflectance, Fig. 3 or Figs. 5&6? These settings would give different reflectance, with the latter being more realistic. This needs to be made very clear in the methodology, including what are these experimental settings used for.
In the tank experiment, what are the optical properties of water? To a minimum, what type of water is that? How the spectra of submersed targets change with depth depend on the water’s optical properties, and therefore it is important to know what water type is used.
How was reflectance calculated from the measurements? Was reflected skylight subtracted? Or is it just a ratio of two measurements (one from the target, either on surface or submerged, and the other from the reference plaque)? This needs to be described with an equation.
Fig. 10. White sail is missing in the caption? Need to explain in the caption. Are these reflectance spectra measured under natural sunlight illumination? Line 225 states that they are from selected depths – but what depths? Why not measuring them at the surface? Also, does each target occupy 100% FOV of the sensor (i.e., equivalent to full coverage of a satellite image pixel)? This is super important because the magnitude of the reflectance in the NIR and SWIR is impacted by the % coverage within the FOV. If half of the FOV is filled with this target debris and the other half being the black background, then the magnitude may be halved. In this regard, why do these spectral magnitudes differ so much in the NIR and SWIR? For example, at 800 nm, the green sail reflectance is < 0.1, but blue foam reflectance is 0.8. This doesn’t appear realistic because most solid materials have similar NIR reflectance magnitude. Are these single measurements or repeated? It’s better to show mean and standard deviation for each type. Are these all dry or wet materials? These questions are important because they will determine whether these spectra can be used as the endmember spectra in interpreting mixed pixels of remote sensing images.
Fig. 11. Apparently these are just one type of gloves and one type of masks. There are other types with different colors and possibly different materials. This needs to be clarified in figure caption.
Fig. 13. Again, why do these spectral magnitudes differ so much in the NIR-SWIR wavelengths, from 0.6 in (a) to 0.02 in (c)? 0.02 is actually approaching the blank reflectance of Fig. 12 – then, is it trustable?
Fig. 13a. Why is the shading area of the dotted purple curve so wide, but there is no shading area in (b) – (f)?
Fig. 14. Why do these spectral magnitudes in the NIR differ so much from those of Fig. 11? Aren’t those the same materials (gloves and masks)?
Supplemental figures: When absorbance is presented, the pathlength or the thickness of the material also needs to be reported. Otherwise the values (not the relative spectral shapes) are meaningless.
Citation: https://doi.org/10.5194/essd-2023-209-CC1 -
RC1: 'Comment on essd-2023-209', Anonymous Referee #1, 23 Jun 2023
This is a nice contribution and I recommend publication. There are few points that could be improved:
The paper provides a nice overview of similar existing open access datasets. However, several experiments are very similar than those presented earlier, so a comparison with previous publications would be an added value for the reader.
Both single pixel (with SEV spectrometer) and multiple pixel observations (with SPECIM hyperspectral camera 400-1000 nm) are performed. I would like to see an intercomparison of both on a pristine sample.
Both indoor and outdoor experiments are performed. Also here, an additional intercomparison of the same samples measured indoor and outdoor would be an added value and would help to interpret the results of the biofouling experiment.
For the indoor experiments, samples are supported by a black aluminum plate. In my understanding, this black plate was not used for the outdoor measurements. Please try to explain what the possible effects of this plate could be.The measurements do not include a correction for the holder, so please advise how to correct for this.
Line 215 and following: increase in reflectance with depth due to lighting geometry? Also the viewing geometry will have an impact as the footprint increases with depth.
Figure 14: Please include a comparison between SEV and SPECIM measurements
Citation: https://doi.org/10.5194/essd-2023-209-RC1 -
RC2: 'Comment on essd-2023-209', Anonymous Referee #2, 30 Jun 2023
This is a well-organized and written paper providing additional measurements of plastic samples, which are an important contribution toward remote sensing of environmental plastic. The two datasets described in this study have similar formats and are available in an open-access repository. The different conditions under which the plastic samples were investigated are well described and compared in the introduction with previous studies of similar nature.
I do recommend publication, I just have a few comments aiming at improving the general quality and flow of the manuscript:
Table 1:This is crucial to illustrate an overview of what are the novelties and strengths of these measurements and compare them with similar research. However, I suggest adding two columns to the table: (i) one column including the conditions in terms of indoor/outdoor experiments and therefore the light source (artificial or natural) and (ii) a second one specifying the type of water used.
Table 2 & Figure 1: It is good to include a full overview of the samples analyzed, including pictures. However, as the polymer types tested in these experiments are already mentioned in Table 1, I think that this table and Figure 1 could be moved to the Annex section.
Line 157-158: repetition of “each panel” at the beginning and the end of the sentence.
Figure 5. This figure illustrates the outdoor experimental setup as well as an overview of the biofouled samples. To reduce the number of images present in the manuscript, and preserve the flow of the paper, I suggest splitting this image in two. Keep in the manuscript the first pictures from a) to e) describing the setup of the outdoor experiments (as Figure 5). The second Figure including the pictures from 01 to 18, showing the biofouled samples could be added to the Annex.
2.2.2 In this section you describe experiments conducted outdoors with measurements taken with natural light. I was wondering if you have information on the sky conditions of the days of the measurements as this might influence your light conditions. In addition, I would make it clearer in this section that the light source is natural light opposed to the one in the laboratory tank experiments.
Results and Discussion section
3.1.1. - 3.1.2 - 3.1.3 and 3.1.4 sections: Although by going back to the material and methods and introduction sections you can retrieve this information, I would suggest clearly stating that these results come from either the tests performed in the tank, and therefore in freshwater and under artificial light, or in the outdoor tank with natural light. I would clarify this within the text and maybe even in the captions of Figures 10 and 11.
Line 225: Can you specify the depths?
Figure 10: At which depths were these measurements taken? Or are these dry measurements, and you additionally have collected the spectral reflectance measurements at selected depths, but they are not reported in the figure? If you can, add the depths within the graph’s legend or in the caption of the figure to clarify. For the white sail, the (a) is missing
Figure 11: The caption needs to be checked as the mask and glove are swapped. (a) is the glove according to the legend of the graph. However, in the caption, it says that a is the medical mask. The same goes for the mask. Graph (b) has "mask" in the legend, but in the caption is "gloves". The (c) is missing the opening bracket.
3.1.4 Blank measurements: are these measurements taken in the tank or the outdoor settings? In other words, what was the experimental setting and therefore the light source used?
3.1.5 section. As with the previous sections, I suggest that it is clearly stated that these measurements were taken in seawater and that natural light was used, as opposed to the artificial light used in the laboratory tank tests.
Figure 13: These measurements are recorded in outdoor conditions, with natural light as a light source. How did you calculate the reflectance? Did you subtract the reflected skylight? Which equation did you use?
Overall this paper is a great effort toward remote sensing of plastic litter, and the datasets presented in this paper are important for the scientific community given the significance of monitoring plastic pollution.
Citation: https://doi.org/10.5194/essd-2023-209-RC2 -
RC3: 'Comment on essd-2023-209', Samantha Lavender, 10 Jul 2023
Thanks for this paper, which will be a valuable addition to supporting the remote sensing of waste plastics.
Having read the feedback from the other reviewers, I feel they've already addressed many of the queries I would have asked. Therefore, I have the following additional comments.
From having undertaken such measurements, I'm missing the number of acquisitions per spectrum presented. To get the final spectrum did you average? If so, for plots such as Figure 10, I would like to see the variation between the duplicate acquisitions. Confidence intervals are mentioned in the legend of Figure 14, but outside of this, I have no information on the measurement variability.
I'm also struggling to see the detail of the spectrum for Figures 7 to 9. Could each figure be followed by a comparison plot, similar to the current Figure 10, showing the surface values for each of the individual panels?
Citation: https://doi.org/10.5194/essd-2023-209-RC3 -
AC1: 'Final response on essd-2023-209', Robin de Vries, 04 Sep 2023
Dear reviewers and community,
We wish to thank you all for taking the time to review our submission. We appreciate the detailed comments and inspiring questions. After an extensive revision, we are confident that we have addressed all comments in the most adequate and practicable way. Due to the large number of comments, we have collected our responses in the attached rebuttal document. We trust that our responses come across most clearly in this manner.
Many thanks once again, on behalf of all authors,
Robin de Vries
Status: closed
-
CC1: 'Comment on essd-2023-209', Chuanmin Hu, 17 Jun 2023
This is a great contribution towards remote detection of plastics in natural environments. The data are made available to the community, which is particularly important to reference against other measurements and to develop algorithms.
For this reason, I recommend publication, but I would also suggest some revisions to improve the presentation.
My biggest concern is how these data can be applied to remote sensing applications. In other words, which spectra can be used as the endmember spectra to interpret remote sensing imagery? Why?
There are also ambiguities here and there that should be corrected or clarified. Please see below.
Leone et al. (2023) did very similar experiments except with a limited depth range of 0 – 0.08 m. So what’s the motivation to do it again? Is the additional depth range of 0.08 – 0.7 m that important? If so, why 0.7 m? Actually both Figs. 8 & 9 show that after 0.08 m, there is nearly no information in the SWIR wavelengths.
For the same reason, how are the experimental results compared with those of Leone et al. (2023) for the same depth range of 0 – 0.08 m? Are they the same or different? If they are different, why?
Same question applies to Knaeps et al. (2021). It’s great to list previous efforts to establish spectral libraries in Table 1 (line 44). But so? Are these earlier efforts not enough? In what way?
What’s the purpose to use take the measurement in Fig. 2? Just to record a photograph of each type?
Fig. 3 shows Halogen tungsten lamps but Figs. 5&6 shows natural illumination. Then, for the reported spectral library, which light source was used? Fig. 3 is not cited but described in the text (line 108 – 117). But what’s the purpose of this experimental setting? My experience with lamps is that they don’t provide collimated beams, meaning that the amount of light received by the target (irradiance) depends on the distance between the target and the light source. Then it will introduce errors when calculating reflectance because the target and the reference plaque may be placed in different positions relative to the light source.
After all, which experimental setting was used to measure reflectance, Fig. 3 or Figs. 5&6? These settings would give different reflectance, with the latter being more realistic. This needs to be made very clear in the methodology, including what are these experimental settings used for.
In the tank experiment, what are the optical properties of water? To a minimum, what type of water is that? How the spectra of submersed targets change with depth depend on the water’s optical properties, and therefore it is important to know what water type is used.
How was reflectance calculated from the measurements? Was reflected skylight subtracted? Or is it just a ratio of two measurements (one from the target, either on surface or submerged, and the other from the reference plaque)? This needs to be described with an equation.
Fig. 10. White sail is missing in the caption? Need to explain in the caption. Are these reflectance spectra measured under natural sunlight illumination? Line 225 states that they are from selected depths – but what depths? Why not measuring them at the surface? Also, does each target occupy 100% FOV of the sensor (i.e., equivalent to full coverage of a satellite image pixel)? This is super important because the magnitude of the reflectance in the NIR and SWIR is impacted by the % coverage within the FOV. If half of the FOV is filled with this target debris and the other half being the black background, then the magnitude may be halved. In this regard, why do these spectral magnitudes differ so much in the NIR and SWIR? For example, at 800 nm, the green sail reflectance is < 0.1, but blue foam reflectance is 0.8. This doesn’t appear realistic because most solid materials have similar NIR reflectance magnitude. Are these single measurements or repeated? It’s better to show mean and standard deviation for each type. Are these all dry or wet materials? These questions are important because they will determine whether these spectra can be used as the endmember spectra in interpreting mixed pixels of remote sensing images.
Fig. 11. Apparently these are just one type of gloves and one type of masks. There are other types with different colors and possibly different materials. This needs to be clarified in figure caption.
Fig. 13. Again, why do these spectral magnitudes differ so much in the NIR-SWIR wavelengths, from 0.6 in (a) to 0.02 in (c)? 0.02 is actually approaching the blank reflectance of Fig. 12 – then, is it trustable?
Fig. 13a. Why is the shading area of the dotted purple curve so wide, but there is no shading area in (b) – (f)?
Fig. 14. Why do these spectral magnitudes in the NIR differ so much from those of Fig. 11? Aren’t those the same materials (gloves and masks)?
Supplemental figures: When absorbance is presented, the pathlength or the thickness of the material also needs to be reported. Otherwise the values (not the relative spectral shapes) are meaningless.
Citation: https://doi.org/10.5194/essd-2023-209-CC1 -
RC1: 'Comment on essd-2023-209', Anonymous Referee #1, 23 Jun 2023
This is a nice contribution and I recommend publication. There are few points that could be improved:
The paper provides a nice overview of similar existing open access datasets. However, several experiments are very similar than those presented earlier, so a comparison with previous publications would be an added value for the reader.
Both single pixel (with SEV spectrometer) and multiple pixel observations (with SPECIM hyperspectral camera 400-1000 nm) are performed. I would like to see an intercomparison of both on a pristine sample.
Both indoor and outdoor experiments are performed. Also here, an additional intercomparison of the same samples measured indoor and outdoor would be an added value and would help to interpret the results of the biofouling experiment.
For the indoor experiments, samples are supported by a black aluminum plate. In my understanding, this black plate was not used for the outdoor measurements. Please try to explain what the possible effects of this plate could be.The measurements do not include a correction for the holder, so please advise how to correct for this.
Line 215 and following: increase in reflectance with depth due to lighting geometry? Also the viewing geometry will have an impact as the footprint increases with depth.
Figure 14: Please include a comparison between SEV and SPECIM measurements
Citation: https://doi.org/10.5194/essd-2023-209-RC1 -
RC2: 'Comment on essd-2023-209', Anonymous Referee #2, 30 Jun 2023
This is a well-organized and written paper providing additional measurements of plastic samples, which are an important contribution toward remote sensing of environmental plastic. The two datasets described in this study have similar formats and are available in an open-access repository. The different conditions under which the plastic samples were investigated are well described and compared in the introduction with previous studies of similar nature.
I do recommend publication, I just have a few comments aiming at improving the general quality and flow of the manuscript:
Table 1:This is crucial to illustrate an overview of what are the novelties and strengths of these measurements and compare them with similar research. However, I suggest adding two columns to the table: (i) one column including the conditions in terms of indoor/outdoor experiments and therefore the light source (artificial or natural) and (ii) a second one specifying the type of water used.
Table 2 & Figure 1: It is good to include a full overview of the samples analyzed, including pictures. However, as the polymer types tested in these experiments are already mentioned in Table 1, I think that this table and Figure 1 could be moved to the Annex section.
Line 157-158: repetition of “each panel” at the beginning and the end of the sentence.
Figure 5. This figure illustrates the outdoor experimental setup as well as an overview of the biofouled samples. To reduce the number of images present in the manuscript, and preserve the flow of the paper, I suggest splitting this image in two. Keep in the manuscript the first pictures from a) to e) describing the setup of the outdoor experiments (as Figure 5). The second Figure including the pictures from 01 to 18, showing the biofouled samples could be added to the Annex.
2.2.2 In this section you describe experiments conducted outdoors with measurements taken with natural light. I was wondering if you have information on the sky conditions of the days of the measurements as this might influence your light conditions. In addition, I would make it clearer in this section that the light source is natural light opposed to the one in the laboratory tank experiments.
Results and Discussion section
3.1.1. - 3.1.2 - 3.1.3 and 3.1.4 sections: Although by going back to the material and methods and introduction sections you can retrieve this information, I would suggest clearly stating that these results come from either the tests performed in the tank, and therefore in freshwater and under artificial light, or in the outdoor tank with natural light. I would clarify this within the text and maybe even in the captions of Figures 10 and 11.
Line 225: Can you specify the depths?
Figure 10: At which depths were these measurements taken? Or are these dry measurements, and you additionally have collected the spectral reflectance measurements at selected depths, but they are not reported in the figure? If you can, add the depths within the graph’s legend or in the caption of the figure to clarify. For the white sail, the (a) is missing
Figure 11: The caption needs to be checked as the mask and glove are swapped. (a) is the glove according to the legend of the graph. However, in the caption, it says that a is the medical mask. The same goes for the mask. Graph (b) has "mask" in the legend, but in the caption is "gloves". The (c) is missing the opening bracket.
3.1.4 Blank measurements: are these measurements taken in the tank or the outdoor settings? In other words, what was the experimental setting and therefore the light source used?
3.1.5 section. As with the previous sections, I suggest that it is clearly stated that these measurements were taken in seawater and that natural light was used, as opposed to the artificial light used in the laboratory tank tests.
Figure 13: These measurements are recorded in outdoor conditions, with natural light as a light source. How did you calculate the reflectance? Did you subtract the reflected skylight? Which equation did you use?
Overall this paper is a great effort toward remote sensing of plastic litter, and the datasets presented in this paper are important for the scientific community given the significance of monitoring plastic pollution.
Citation: https://doi.org/10.5194/essd-2023-209-RC2 -
RC3: 'Comment on essd-2023-209', Samantha Lavender, 10 Jul 2023
Thanks for this paper, which will be a valuable addition to supporting the remote sensing of waste plastics.
Having read the feedback from the other reviewers, I feel they've already addressed many of the queries I would have asked. Therefore, I have the following additional comments.
From having undertaken such measurements, I'm missing the number of acquisitions per spectrum presented. To get the final spectrum did you average? If so, for plots such as Figure 10, I would like to see the variation between the duplicate acquisitions. Confidence intervals are mentioned in the legend of Figure 14, but outside of this, I have no information on the measurement variability.
I'm also struggling to see the detail of the spectrum for Figures 7 to 9. Could each figure be followed by a comparison plot, similar to the current Figure 10, showing the surface values for each of the individual panels?
Citation: https://doi.org/10.5194/essd-2023-209-RC3 -
AC1: 'Final response on essd-2023-209', Robin de Vries, 04 Sep 2023
Dear reviewers and community,
We wish to thank you all for taking the time to review our submission. We appreciate the detailed comments and inspiring questions. After an extensive revision, we are confident that we have addressed all comments in the most adequate and practicable way. Due to the large number of comments, we have collected our responses in the attached rebuttal document. We trust that our responses come across most clearly in this manner.
Many thanks once again, on behalf of all authors,
Robin de Vries
Robin V. F. de Vries et al.
Data sets
Dataset of spectral reflectances and hypercubes of submerged biofouled, pristine, and ocean-harvested marine litter Robin V. F. de Vries, Shungudzemwoyo P. Garaba , Sarah-Jeanne Royer https://doi.org/10.4121/7c53b72a-be97-478b-9288-ff9c850de64b.v1
Dataset of spectral reflectances and hypercubes of submerged plastic litter, including COVID-19 medical waste, pristine plastics, and ocean-harvested plastics Robin V. F. de Vries, Shungudzemwoyo P. Garaba https://doi.org/10.4121/769cc482-b104-4927-a94b-b16f6618c3b3.v1
Robin V. F. de Vries et al.
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
406 | 86 | 20 | 512 | 5 | 8 |
- HTML: 406
- PDF: 86
- XML: 20
- Total: 512
- BibTeX: 5
- EndNote: 8
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1