Consistency between Satellite Ocean Colour Products under High Coloured Dissolved Organic Matter Absorption in the Baltic Sea

Round 1

Reviewer 1 Report

The manuscript provides an assessment of the radiometric products from various satellite ocean color missions for the Baltic optically complex waters. The assessment relies on in situ data from AERONET-OC stations and shipborne measurements. Review provided by G. Zibordi. 

The manuscript is suitable for publication in Remote Sensing assuming the following comments are considered.

Comment 1. Page 2 “…, due to limitation of the method of applying bi-directional correction factors …”. AERONET-OC data are provided at all center-wavelengths including those in the NIR. BRDF corrections are simply not applied at center-wavelengths beyond 709 nm (see the paper by Zibordi et al. 2021 in JAOT). The fact that autonomous shipborne spectro-radiometers may fill spectral gaps, is at least not precise. Since 2018 the new AERONET-OC radiometers have an extended number of bands matching those of OLCI. In addition, if NIR data are considered relevant for matchup analysis in the Baltic waters, it should be mentioned which corrections are (or should be) applied. The above elements require clarifications.

Comment 2. Page 2. References on AERONET-OC should include the JAOT paper of 2021.

Comment 3. Page 4. “The standard deviation of 3x3 pixels was used as a metric of uncertainty for the satellite data”. The standard deviation of the 3x3 pixels simply provides the variability across the 3x3 elements. It is just one of the contributions to the uncertainty. It is speculative indicating or assuming it can be a metric for the uncertainty. The text should be revised.

Figure 3. Pag. 5. AERONET-OC and ship data should be identified with different colors. They were determined using diverse instruments and methods. This may imply systematic differences across the two data sets.

Page 6. Figure 4 and Table 2. It should be clarified how many images contributed to the matchups constructed utilizing ship data, and how these matchups were actually constructed (e.g., which is the minimum distance between matchups in each in situ transect when referring to the same satellite image). In addition, AERONET-OC instruments operated in the Baltic do not have the 709 nm center-wavelength. Still, there are 709 nm data associated to matchups (see the crosses in the last panel in Fig. 3 and the identical number of matchups associated to each spectral band in Table 2). This requires clarification.

Page 6 and 7. Table 2. Some of the center-wavelengths (e.g., 560 nm) were only available from 2018. Was there any correction applied to AERONET-OC data?

Page 9. Discussion session. This session looks more a review than an actual discussion. It would be interesting to have some guess on the reasons for the diverse performance of the various atmospheric corrections. Just some general hypothesis.

Author Response

Reviewer #1 Comments.

General Comment. The manuscript provides an assessment of the radiometric products from various satellite ocean color missions for the Baltic optically complex waters. The assessment relies on in situ data from AERONET-OC stations and shipborne measurements. The manuscript is suitable for publication in Remote Sensing assuming the following comments are considered.

Response: We thank Reviewer #1 for their appraisal of this work. We have addressed all of the comments raised. A response to each comment is given below.

Comment 1. Page 2 “…, due to limitation of the method of applying bi-directional correction factors …”. AERONET-OC data are provided at all center-wavelengths including those in the NIR. BRDF corrections are simply not applied at center-wavelengths beyond 709 nm (see the paper by Zibordi et al. 2021 in JAOT). The fact that autonomous shipborne spectro-radiometers may fill spectral gaps, is at least not precise. Since 2018 the new AERONET-OC radiometers have an extended number of bands matching those of OLCI. In addition, if NIR data are considered relevant for matchup analysis in the Baltic waters, it should be mentioned which corrections are (or should be) applied. The above elements require clarifications.

Response to Comment 1: We have modified the text accordingly and added the Zibordi et al. 2021 JAOT reference. The sentence: ‘AERONET-OC  is also not available at all wavebands, due to the limitation of the method of applying bi-directional correction factors to bands in the red and NIR [29].’ has now been deleted.

The text on page 2 now reads: The addition of a new suite of AERONET-OC radiometers since 2018, has extended the number of spectral bands from 9 to 12 to match those available from OLCI [28b].’

In addition on page 3 (lines 92-98), we provide updated details of the new AERONET-OC radiometers as follows: ‘Since 2018, the deployment of CE-318T 12-channel radiometer systems within the AERONET-OC network has provided additional spectral bands centred at 400, 510, 620 and 779 nm at marine water sites and 681 and 709 nm for inland water sites. The CE-318T systems also increased the frequency of measurement sequences from 30 minutes to 5 minutes to provide a finer temporal scale resolution to detect environmental perturbations’. We did not use the CE-318T data in this analysis. The NIR data for the AERONET-OC sites were interpolated from red wavelengths.

For NIR corrections on the ship data, Individual spectra were inspected to evaluate the shape of the NIR signal due to high particle scattering. When no particle scattering were observed, any NIR offset was corrected for by subtracting the mean Rrs in the near infrared region from 850–900 nm (Qin et al. 2017). To clarify this, the following sentences have been added on page at lines 97 to 105: ‘It is generally assumed in waters with low particle scattering that NIR reflectance is close to zero (Hooker et al., 2002). This assumption generally holds in the Baltic Sea outside peak productivity periods or close to rivers and shallow areas, when there higher concentrations of phytoplankton, detrital material or sediment in surface waters may be present. Additionally, residual surface water effects such as spray, sun glint and whitecaps will elevate  in the visible and NIR. Individual spectra were inspected to evaluate the shape of the NIR signal for signs of high particle scattering. When no elevated particle scattering was observed, as evidenced by a spectrally flat NIR signal, any offset observed in the NIR was assumed to be caused by spectrally neutral effects and corrected for by subtracting the mean  from the entire spectrum (Qin et al. 2017).’

Comment 2. Page 2. References on AERONET-OC should include the JAOT paper of 2021.

Response to Comment 2: The reference has been added.

Comment 3. Page 4. “The standard deviation of 3x3 pixels was used as a metric of uncertainty for the satellite data”. The standard deviation of the 3x3 pixels simply provides the variability across the 3x3 elements. It is just one of the contributions to the uncertainty. It is speculative indicating or assuming it can be a metric for the uncertainty. The text should be revised.

Response to Comment 3: This sentence has been removed as suggested.

Comment 4. Figure 3. Page 5. AERONET-OC and ship data should be identified with different colors. They were determined using diverse instruments and methods. This may imply systematic differences across the two data sets.

Response to Comment 4: The figure has been modified as suggested and discussion of the different in situ data has been added to page 20, lines 408 – 430 and page 23, lines 536 - 550.

Comment 5. Figure 4 and Table 2. Page 6. It should be clarified how many images contributed to the matchups constructed utilizing ship data, and how these matchups were actually constructed (e.g., which is the minimum distance between matchups in each in situ transect when referring to the same satellite image). In addition, AERONET-OC instruments operated in the Baltic do not have the 709 nm center-wavelength. Still, there are 709 nm data associated to matchups (see the crosses in the last panel in Fig. 3 and the identical number of matchups associated to each spectral band in Table 2). This requires clarification.

Response to Comment 5: We apologies if this was not clear. The in situ data (1-min bins) were matched to individual satellite pixels. From the 3 x 3 pixels, the centre pixel was used for the validation procedure. The in situ data (1-min bins) were matched to individual satellite pixels. All in situ data within a specific pixel were averaged, so that each match-up has an independent set of in situ data and there was no overlapping in situ data between match-ups. The minimum distance between matchups in the satellite image is therefore reflected in the resolution of each sensor which for OLCI-A is 300m, MODIS-Aqua is 1km and VIIRS is 750 m. The validation statistics were computed on the centre pixel, to ensure that each match-up uses an independent satellite pixel. Additionally, the standard deviation around the match-up (from the 3 x 3 box) was computed as an index of the homogeneity of the match-up. The number of granules for each sensor were 38 for OLCI-A, 40 for MODIS-Aqua and 7 for VIIRS. This has now been added to the methods section.

Data from the CE-318 9-channel radiometer systems were used. The NIR band at 709 nm, which was linearly interpolated from the 665 nm band.

Comment 6. Table 2. Page 6 and 7. Some of the center-wavelengths (e.g., 560 nm) were only available from 2018. Was there any correction applied to AERONET-OC data?

Response to Comment 6: Yes the AERONET-OC data were band shift corrected following Zibordi et al. (2009a). Apologies for missing this detail which has now been added to the revised version on page 2 from lines 96-101 as follows: ‘The AERONET-OC waveband centres are 413, 441, 491, 555, 668 and 870 nm in HLT, and 412, 439, 500, 554, 675 and 870 nm in GDLT; To match with the spectral bands of OLCI-A, MODIS-Aqua and VIIRS, water-leaving radiance corrected for viewing angle dependence and the effects of the non-isotropic distribution of the in-water radiance field (-f/Q) were band shift corrected following Zibordi et al. [28], based on a regional bio-optical algorithm to reduce inter-band uncertainties. The corresponding  were computed from the band-shifted -f/Q data using the extra-atmospheric solar irradiance () for each waveband as in Qin et al. [30].

Comment 7. Discussion session. Page 9. This session looks more a review than an actual discussion. It would be interesting to have some guess on the reasons for the diverse performance of the various atmospheric corrections. Just some general hypothesis.

Response to Comment 7: The Discussion has been re-written to include possible reasons for the performance of the various AC processors (see pages 21-23).

Reviewer 2 Report

It’s a big challenge to do AC for waters with complex optical properties. Data users also want to know the consistency between different satellite ocean color products. This paper provides us much valuable information about this topic for waters with high CDOM absorption contribution in the Baltic Sea. A powerful in situ database was used for assessing the accuracy of existing ocean color products, which displayed the better performance of POLYMER over other AC processors for OLCI-A and the underestimation of MODIS-Aqua and Suomi-VIIRS. Moreover, it also expands the analysis to the comparison of spatial distribution of Rrs value at specific wavebands.

I recommend the publication of this manuscript after moderate revisions.

1. Do we need to add the acronym for the bias-corrected root-mean-square-error at Section 2.3, as well as in Table2.
2. Detailed figure caption for Figure 5 is needed. It’s not so clear.
3. Figures are not clear enough. Please clarify them.
4. In the discussion, the authors mainly focused on the comparison results of different products, but less on the reasons attributing these departures. More explanations and deep discussion about it are expected. I hope this study could provide referential value for other sea areas with similar optical properties.

Author Response

Response to Reviewer #2.

Reviewer #2 Comments.

General Comment. It’s a big challenge to do AC for waters with complex optical properties. Data users also want to know the consistency between different satellite ocean color products. This paper provides us much valuable information about this topic for waters with high CDOM absorption contribution in the Baltic Sea. A powerful in situ database was used for assessing the accuracy of existing ocean color products, which displayed the better performance of POLYMER over other AC processors for OLCI-A and the underestimation of MODIS-Aqua and Suomi-VIIRS. Moreover, it also expands the analysis to the comparison of spatial distribution of Rrs value at specific wavebands.

I recommend the publication of this manuscript after moderate revisions.

Response to General Comment: We thank Reviewer #2 for their appraisal of this work. We have addressed all of the comments raised, as highlighted below.

1. Do we need to add the acronym for the bias-corrected root-mean-square-error at Section 2.3, as well as in Table2.

Response to Comment 1: The acronym for the bias-corrected root-mean-square-error at Section 2.3, as well as in Table2 has now been corrected.

1. Detailed figure caption for Figure 5 is needed. It’s not so clear.

Response to Comment 2: Further details have been added to the figure caption for Figure 5 as follows: Composite (λ) images for each AC for the Baltic Sea for the period from 11 to 17 June 2016 are shown on the left and right panels. On the left panel, from top to bottom the images are for; OLCI-A pb 2, POLYMER v4.13 and C2R-CC v2.0. The right panel, from top to bottom, the images are for OLCI-A OL_L2M.003.00, VIIRS R2018 and MODIS-Aqua R2018. Centre panels are comparison of (λ) at 443, 560 and 665 nm for each AC along two transects shown as red lines in the top left image. The data were extracted at every 20 km for each AC model. The centre left data are (λ) for the different ACs extracted from the north-south transect from the Gulf of Bothmia to the Baltic proper (shown in the top left pb2.23-2.29 image). The centre right data are (λ) for the different ACs for the east-west transect from the head of the Gulf of Finland through the HLT to the GDLT. For these transects, The black points are in situ (λ), yellow triangles are OLCI-A pb2.23-2.29, red circles are OL_L2M.003.00, purple squares are OLC-A C2R-CC, grey triangles are OLCI-A POLYMER v4.13, green circles are VIIRS and blue squares are MODIS-Aqua.

1. Figures are not clear enough. Please clarify them.

Response to Comment 3: The figures have been modified as suggested. For Figure 3, the AERONET-OC and ship data have now been separated in different panels (as also suggested by Reviewer #1). To make Figure 4 clearer, rather than showing all six AC processors in one figure, which we agree can obscure the salient trends, two ACs are now plotted per figure. There are therefore now three scatter plots instead of one. The Figure legends for the original Figures 4 and 5 have also been modified to help to clarify these Figures. To aid readability of the ms, the corresponding statics for each of the new figures have been placed directly below them. In addition, the areas of the Baltic Sea, reffered to in the text, have now been labelled in Figure 1. We hope that this makes the figures clearer.

1. In the discussion, the authors mainly focused on the comparison results of different products, but less on the reasons attributing these departures. More explanations and deep discussion about it are expected. I hope this study could provide referential value for other sea areas with similar optical properties.

Response to Comment 4: The Discussion has been re-written to include reasons for the trends in the AC processors that are presented (see pages 21-23). Reference to other seas with similar optical properties has been added on page 21, at lines 479-482.

Reviewer 3 Report

Title: Consistency between satellite ocean colour products under high coloured dissolved organic matter absorption in the Baltic Sea

This manuscript evaluated four atmospheric correction processors for OLCI-A and standard processors for MODIS-Aqua and VIIRS by using in situ remote sensing reflectance spectra data which include the shipborne measurements and two AERONET-OC sites. The assessment was based on the objective statistical comparisons of in situ and satellite data from April 2016 to September 2018 and suggested that the most suitable processor for OLCI-A is POLYMER. Similar results are observed in the standard processors for MODIS-Aqua and VIIRS during 11-17 June 2016. 

Major Comments:

As an end-user, I appreciate that the authors provided thorough assessments for these remote sensing algorithms; however,  the majority of the explanation provided for the statistical results (Table 2) focus on the relative percentage difference, please provide a more thorough interpretation of Table 2.   

Minor Comments:

1. The title is “Consistency between satellite ocean colour products under high coloured dissolved organic matter absorption in the Baltic Sea”, however, the conclusions suggest that OLCI POLYMER will generate the most accurate biogeochemical monitoring water quality parameters. Does that mean the level of consistency is low or high?
2. How does the seasonal river discharge vary the TSM and CDOM concentrations and further change the R_rs? 
3. There are a few typos and missing symbols such as “opean-ocean”, “486 560”, “root-mean-square error ( )”.
4. The Figure caption of Figure 4 does not provide the information about the star marker.

Author Response

Response to Reviewer #3.

Reviewer #3 Comments.

General Comment.

As an end-user, I appreciate that the authors provided thorough assessments for these remote sensing algorithms; however,  the majority of the explanation provided for the statistical results (Table 2) focus on the relative percentage difference, please provide a more thorough interpretation of Table 2.

Response to General Comment: The results section has been completely re-written to provide a more thorough explanation of the statistical results given in Table 2. See pages 7 and 8. We hope the Reviewer now finds this satisfactory.

Minor Comments:

1. The title is “Consistency between satellite ocean colour products under high coloured dissolved organic matter absorption in the Baltic Sea”, however, the conclusions suggest that OLCI POLYMER will generate the most accurate biogeochemical monitoring water quality parameters. Does that mean the level of consistency is low or high?

Response to Comment 1: The level of consistency is low (as seen in Figures 3, 4 and 5). From the results we suggest that the AC POLYMER is applied to each sensor (OLCI-A, MODIS-Aqua and VIIRS) to provide a consistent long time series data set. This has now been qualified in the discussion and conclusion sections as follows, lines 292-293 now read: ‘In this study, we found that the consistency between different OLCI-A, MODIS-Aqua and VIIRS AC products is poor’. Line 300-301 now reads: ‘To improve the consistency between OLCI-A, MODIS-Aqua and VIIRS, the application of the POLYMER AC to these sensors should be tested and validated’. We have also added a similar statement to the end of the Abstract.

2. How does the seasonal river discharge vary the TSM and CDOM concentrations and further change the R_rs? 

Response to Comment 2: The variability in CDOM and TSM at river mouths, is expected to decrease (λ) at blue and blue-green bands as the ratio of backscatter to absorption increases, which would produce a higher slope in the (λ) spectra from the blue to the green. An increase in TSM loads and therefore backscatter would also be observed in an increase in the offset in (λ) in the NIR. For the match-ups in this study however, the influence of seasonal river discharge is low as the ferry tracks mostly traverse the deeper waters of the Baltic Sea and the AERONET-OC sites are located away from major rivers. This has been added to the Discussion section on lines 401-407.

3. There are a few typos and missing symbols such as “opean-ocean”, “486 560”, “root-mean-square error ( )”.

Response to Comment 3: The typos and missing symbols have now been corrected; see lines 26, 43 and 166.

4. The Figure caption of Figure 4 does not provide the information about the star marker.

Response to Comment 4: The star symbol represents data from the AERONET-OC site HLT. To make Figure 4 clearer, rather than showing all six AC processors in one figure, which can obscure the salient trends and the star points, two ACs are now plotted per figure. There are now three scatter plots instead of one to facilitate clearer visibility of the star points.