the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Jan 2021
Radiative transfer modeling with BGC-Argo float data in the Mediterranean Sea
Abstract. A radiative transfer model was parameterized and validated using Biogeochemical Argo float data acquired between 2012 and 2017 across the Mediterranean Sea. Fluorescence-derived chlorophyll a concentration, particle backscattering at 700 nm and fluorescence of colored dissolved organic matter were used to parametrize the light absorption and scattering coefficients of the optically significant water constituents (pure water, non-algal particles, colored dissolved organic matter and phytoplankton). The model was validated with in-situ downwelling irradiance profiles and irradiance-derived apparent optical properties from satellite data, such as the diffuse attenuation coefficients and remote sensing reflectance. To the authors' knowledge, this is the first time that a three-platform comparison of such kind is performed between model, floats and satellites. Results showed that by using regional parameterizations that are not only related to chlorophyll concentration and vertical distribution, the model was able to capture a more accurate spectral response in the examined wavelength range compared to chlorophyll-related (or Case 1) optical models. When using alternative models that incorporated also measurements of colored dissolved organic matter fluorescence or particulate optical backscattering, the model skill increased at all examined wavelengths. A series of upgrades, such as the inclusion of temperature and salinity data for the modification of the pure water absorption spectra, a refined pure water absorption model, as well as the correction of regional algorithms that had overestimated the pure water contribution in the blue, all contributed to improve the model performance. Finally, using a multi-spectral optical configuration enabled to estimate also the relative contribution of separate water constituents in the examined spectral range. Simulations including non-algal particles and colored dissolved organic matter performed up to 60 % and 76 % better than when considering the optical properties of pure seawater alone. Moreover, a simulation including phytoplankton absorption resulted in an error reduction of up to 43 %, especially at 412 nm and with a more uniform response at the wavelengths considered. Such studies can therefore also tackle the bio-optically anomalous nature of the Mediterranean Sea, and show that non-chlorophyll-related constituents (i.e. non-algal particles and colored dissolved organic matter) can substantially modulate the underwater light field in the blue.
This preprint has been withdrawn.
-
Withdrawal notice
This preprint has been withdrawn.
-
Preprint
(5268 KB)
Interactive discussion
Status: closed
-
RC1: 'Comment on bg-2020-473', Emmanuel Boss, 15 Feb 2021
Review of: ‘Radiative transfer modeling with BGC-Argo float data in the Mediterranean Sea’ by Terzic et al.
Reviewer: Emmanuel Boss
This paper studies the utility using simple radiative transfer (RT) model + input from BGC-Argo floats to obtain reasonable optical properties in the water column. A variety of different models are used and evaluated using radiometry and remote sensing measurements.
The topic of the paper is novel and of interest to the biogeosciences readership. I think that with appropriate revisions it could be a useful contribution.
I have major comments that I detail here and more minor one that are on the PDF I am sending back.
- The paper needs some more editing to be easily readable.
- Its organization could be improved a lot using tables describing the different configuration used (e.g. Which specific IOP choice for each, which property profile is used for each) and to display the results.
- The neglect of Raman needs significant justification, in particular at 490nm where it could significantly affect measured Kd and Rrs.
- The neglect of a_phi at 380nm deserves more justification. Works of Bricaud, among others, suggest significant phytoplankton absorption there due to MAAs, particularly near the surface.
- The effect of S and T on water IOPs is very different in the bands investigated. S primarily contribute to backscattering but if a mean salinity is used, the change may be rather small. T affects primarily absorption in the NIR. Again, using an average T may be more than sufficient anywhere else.
- The final result that the best choice is bbp_tilde =0.002 and eta=0 deserves attention. It does not seem consistent with expectation from other studies.
- It is not clear why there seem to be no configuration with Fchl providing the vertical profile for a_phi, Fdom for a_cdom and b_bp for b_bp and NAP. If this one does not work well, please explain why you think it does not? What are the implications?
Dear authors: I am often wrong. If you feel my comments are ‘off mark’ please contact me and if convinced I will be happy to change them.
-
AC1: 'Reply on RC1', Elena Terzić, 09 Mar 2021
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2020-473/bg-2020-473-AC1-supplement.pdf
-
RC2: 'Comment on bg-2020-473', Anonymous Referee #2, 15 Feb 2021
This manuscripts presents a radiative transfer model that uses BGC-Argo float data to derive profiles of the underwater light field. This is a thorough, comprehensive study which tested different inherent optical property models/parameterizations, and validated them on float-measured downwelling irradiance and satellite observations.
Given the increasing use of BGC-Argo floats and autonomous gliders, methods to connect their biogeochemical measurements to optical properties/underwater light fields are important – this study is one of few attempting to fill that gap, and as such, is of interest to the wider community.
Ultimately, I think this manuscript could be a valuable contribution, however in its current form, I have some concerns/comments and areas that I think need a bit of clarification.
1) Some of the method details and corresponding results are a bit hard to follow. Sometimes it’s unclear what exact IOP parameterization is used where, and what parameterization the different results in the text are referring to. I think this could be simplified with a table or two summarizing these details.
2) Maybe I’m missing something in your methodology completely, but it is unclear to me why you test the individual OSM (optically significant material) constituent IOP models separately. You know that for your float data, you’re never in pure water or pure water + 1 OSM. You’re almost always going to be in your 6th IOP model group (i.e. all constituents). So you need to test your different individual OSM-IOP model parameterizations within this 6th IOP model group because it is the total IOPs that are important. For example, when testing your a_NAP model parameterizations (your 2nd IOP model group), you may get a better result (i.e. an irradiance that is closer to the measured value) for one of the parameterizations that is actually less correct. This is because that parameterization could be less correct in a way that is more correct in terms of the actual total IOP e.g. in this NAP case, perhaps your NAP model is incorrect, but it accurately captures CDOM, and CDOM is the dominant OSM in that sample, so overall, the total absorption is more correct.
Further comments are detailed below.
Specific comments
L7: Lefering et al (2020) used model, satellite and glider data for the type of three-platform comparison done here. [Development of bio-optical model for the Barents Sea to quantitatively link glider and satellite observations, Philosophical Transactions A: Mathematical, Physical and Engineering Sciences https://doi.org/10.1098/rsta.2019.0367]
Fig 1 & 2: units missing on the colour bar – number of floats? Number of days with profiles present over the 5 year period?
L75: you haven’t defined Kd yet
L75: The details on the Kd calculation are not clear. It seems like you are doing a depth averaged Kd rather than taking the derivative of the Ed profile (to get a profile of Kd). If so, what depth are you averaging over? Why do you need to have the first Ed measurement shallower than 1m?
L78: Are you saying that you tested all your different model parameterizations but then just removed the results that had an Ed difference of greater than 30%? What’s the justification for this?
L95, 99: missing backscattering coefficients
L142: What is the “this” at the start of the sentence “This is however present…” Can you please reword to make it clearer?
Eq 16 / L177: I don’t understand this. Why are you estimating aCDOM using Pope and Fry when you are using Mason et al elsewhere?
L178-179: in the text you have a_{ORIG}, but in Eq 16 you have a_w^{ORIG} – can you please make it consistent?
L195: No other OSMs affecting Kbio(380)? Can you expand on this in the text please.
Section 2.3.4: How do the min and max Chl values affect the results? You didn’t see any features/performance issues at the low and high ends of your data?
Eq 20: using chl as a bp model – what about NAP?
Are all IOP models derived for the “surface” and then extrapolated? What’s the “surface”?
Fig 6 & L310: what configurations are used for each of the OSMs here? The best aNAP and aCDOM results from Figs 4 and 5? You described a range of different scattering models – which one is shown in this figure?
L311-312: Suggest including that these decreases in RMSE and bias are in reference to the pure water simulation.
L323-326: what run is this paragraph referring to? I’m assuming the run with water + all OSMs. If so, what configurations are used for each of the OSMs? Can you please include those details?
Fig 7: The y-axis label isn’t defined anywhere
Fig 8 – 10: what is the depth bin or range for the model and float data shown in these figures?
I’m not sure I’m following the aNAP argument presented in section 3.2. You state the Kd comparisons improve when you remove the aNAP component – this makes sense because of your assumption that Kbio was only CDOM driven. Then you say you need the aNAP component to get the best Rrs retrievals, but offer no explanation. What’s going on here? It seems your Rrs model is inconsistent with your Kd model?
Technical corrections, typos etc
L10 “also incorporated” rather than “incorporated also”
L15 “also enabled the estimation of” rather than “enabled to estimate also”
L27 Suggest reordering, the second part of the sentence is fragmented. “To achieve this, and in order to follow up with the pace… in-situ platforms, the implementation…remains essential.”
L49-50: Same as above, suggest reordering “With their high vertical resolution profiles and high spatial coverage, BGC-Argo…”
L125: Haven’t defined PFT
L135 and a few other places: you have an indent after the equation where there shouldn’t be one.
L141 “of conveying” rather than “to convey”
L163 the Babin et al ref should be in parenthesis
L195-199: This sentence is slightly confusing. I think you’re saying the IOPs used for pure water are an upgrade because you use Mason et al and include a T-S correction. Suggest rewording along the lines of “Given the fact that the IOP models used here for pure water absorption are the new measurements of Mason et al, and a T-S correction is applied, …”
Eq 12 and 22: use bb’ and bb-tilde for backscattering ratio, please make it consistent throughout.
Fig 3: In the caption the formatting of the xlabels is different from the figure.
Fig 5 Caption: “twe” should be “the”
Citation: https://doi.org/10.5194/bg-2020-473-RC2 -
AC2: 'Reply on RC2', Elena Terzić, 09 Mar 2021
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2020-473/bg-2020-473-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Elena Terzić, 09 Mar 2021
-
RC3: 'Comment on bg-2020-473', Zach Erickson, 16 Feb 2021
The authors use BGC-Argo data to compare IOPs (absorption and scattering) and AOPs (Kd and Rrs) estimated using simple water column radiative transfer, in situ measurements of water column constituents, and satellite remote sensing. The focus of this project is the Mediterranean, where the optically complex waters justify comparing different approaches. This paper directly compares different estimates of absorption spectra, and in many cases shows the utility both of more recent estimates of a_w, as well as the benefit of a regional PFT approach for pigment absorption.
While this research is definitely useful, in many places I struggled to understand what exactly the authors were doing. In particular, in Section 2.3 they list 6 different models that they test, comparing measured downwelling irradiance with that derived from different models following Eq. 1-12. Their first model is pure water absorption and scattering, for which they test three different estimates of a_w (Fig. 3). It wasn’t clear to me what downwelling irradiance observations they used to test the different a_w. From the text, it seems like they compared actual profiles of Ed to estimated profiles of Ed assuming clear water, but that doesn’t make any sense. Did they only use BGC-Argo profiles in very clear waters, and if so, how did they define that? Or did they have a “basic” version of the other IOPs that they used in Eq. 1-12, in which case, what was it, and how does it differ from experiment #6? I have a similar question for experiments 2-5; from the text it seems like the authors assume the only constituents are those specifically tested, but that doesn’t make sense unless you have measurement profiles for which that might be true. I’m sure I am missing something here.
Another point of confusion was how the authors use estimates of Chl from BGC-Argo. BGC-Argo floats do not measure Chl, they measure fluorescence from Chl (fChl). During the daytime, near the surface fChl can be reduced even though Chl is still elevated because of non-photochemical quenching (NPQ). There are a number of methods to correct for this, including specifically for BGC-Argo (e.g., Xing et al., 2018). Was any sort of NPQ correction applied to the fChl data? If not, I would be worried that the authors’ “Case I models” are tracking fChl and not Chl.
Finally, I was uncertain about how to interpret Figures 8-10 in light of the text on l. 254-256 and 332. Are the satellite values shown monthly climatological values? If so, I don’t see how these can be compared with Argo considering the time frames and spatial locations are different (that is, Argo doesn’t sample uniformly and because of cloud cover neither does the satellite, and they also have different temporal weights based on the different lifetimes of the satellites and Argo floats).
Figures in general: I recommend removing Figures 1 and the top panel of 2 (see comment below), and moving the information in Figures 3-6 into a single Table, after removing experiments that don’t make sense to include (see comments below). What would be more useful to see in a Figure would be some representative profiles of measured Chl, bbp, Ed, and modeled Ed. Instead of Figure 7, what would be more useful to see would be the regionality or where the measurements are close to or far away from the 1:1 line with the model.
Conclusions Section in general: A lot the text in the Conclusions does not logically follow from the results of this paper. For example, I don’t see a clear path between this paper and combining oxygen, nitrate, and pH in numerical models (l. 412-417), and I’m not convinced that this paper in particular demonstrates the need for inclusion of multi-spectral measurements (l. 423-424) or hyperspectral models (425-427). To be clear, I don’t necessarily disagree with any of these statements, I just don’t think these are conclusions or logical next steps that one would arrive at from reading this paper.
Smaller comments:
I don’t think it is useful to show how many BGC-Argo profiles didn’t have the right set of measurements at the right depths to do this analysis. If profiles were discarded because of data quality (rather than data availability) that could be useful to know – e.g., “a condition of less than 30% difference between modelled and computed Ed values was thus added which resulted in 147 profiles less” (l. 77-78), although I don’t quite understand what this means – but right now Figure 1 and the top panel of Figure 2 don’t add anything to the paper.
Sec. 2.2 could use more textual help for the equations – make sure to define variables in the same paragraph where they first appear and provide text to explain the equations. Also, when denoting variables associated with direct downward irradiance sometimes a subscript of “dir” is used (e.g., E_dir) and sometimes “d” (e.g., C_d); please pick one. Similarly for “dif” and “s”. Please also check in this section and throughout to make sure b and b_p mean scattering, and b_b and b_bp mean backscattering, and that variable dependencies are correct (e.g., a_w doesn’t depend on z).
l. 125: Define PFT.
The text around Eq. 13 is confusing; I don’t think this equation is actually helpful. The authors can just say in the text in Sections 2.3.3 and 2.3.4 that profiles either follow Chl, b_bp, or fDOM.
l. 141-142: I am confused by the statement that AOPs “to a certain extent remove the impact of the external environment’s variability”; by definition, AOPs are properties that depend on these external parameters.
Sec. 2.3: I found some of this section backwards – the text in l. 194-195 would be nice to have before the authors explain the text around Eq. 16. Also, considering the authors have chosen to use Mason et al. 2016 instead of Pope and Fry (1997), why show the test with the Pope and Fry-derived values at all? Similarly, in Fig. 5, why show the values without the T/S correction for a_w?
l. 229-230: If the estimated variability in this backscattering spectral power-law slope parameter is from 0-4, using a constant value of 2 seems overly simplistic. Some discussion about error and uncertainty as a result of this would be good to include.
l. 233: What final value of the backscattering ratio was used? What error and uncertainty is the result?
l. 336-337: The authors could be more explicit about what the two Kd estimates actually are here (both are float-derived). There should also be more discussion somewhere – why might IOPs be overestimated?
Need a statement about data availability somewhere (typically in Acknowledgements). Also, should specifically list which BGC-Argo floats were used.
Minor edits (not exhaustive):
l. 157-158: I think it is more correct to say that the absorption spectrum is often modeled (not “follows”) as an exponentially decreasing shape despite its heterogeneous biogeochemical composition.
l. 173: “expressed” should be “parameterized”? Similarly, in l. 225 “obtained” should be “estimated”.
l. 201-202: Organelli et al. (2017c) doesn’t provide spectra for wavelengths shorter than 400 nm; how do you get data at 380 nm? Also, how representative are these of Mediterannean waters? And finally, some explicit response should be given to that paper’s caveat that their results “were not intended for any algorithm development and/or validation” (to quote from their Conclusions).
l. 213-214: How many data points have Chl outside of these bounds? Would it not be preferable to just omit those profiles? If the authors keep them in, some discussion should be present about what error this introduces.
Eq. 21: negative sign missing in the exponent?
Fig. 5: Is it still considered Case I if the CDOM profile follows fDOM and not Chl?
l. 368: It should be noted that satellite-derived Rrs(412) is prone to large uncertainty, especially in optically complex waters; see e.g. Wei et al. (2020) and references therein.
Fig. 7: I don’t think the word BIOPTIMOD was anywhere in the main text?
Citation: https://doi.org/10.5194/bg-2020-473-RC3 -
AC3: 'Reply on RC3', Elena Terzić, 09 Mar 2021
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2020-473/bg-2020-473-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Elena Terzić, 09 Mar 2021
Interactive discussion
Status: closed
-
RC1: 'Comment on bg-2020-473', Emmanuel Boss, 15 Feb 2021
Review of: ‘Radiative transfer modeling with BGC-Argo float data in the Mediterranean Sea’ by Terzic et al.
Reviewer: Emmanuel Boss
This paper studies the utility using simple radiative transfer (RT) model + input from BGC-Argo floats to obtain reasonable optical properties in the water column. A variety of different models are used and evaluated using radiometry and remote sensing measurements.
The topic of the paper is novel and of interest to the biogeosciences readership. I think that with appropriate revisions it could be a useful contribution.
I have major comments that I detail here and more minor one that are on the PDF I am sending back.
- The paper needs some more editing to be easily readable.
- Its organization could be improved a lot using tables describing the different configuration used (e.g. Which specific IOP choice for each, which property profile is used for each) and to display the results.
- The neglect of Raman needs significant justification, in particular at 490nm where it could significantly affect measured Kd and Rrs.
- The neglect of a_phi at 380nm deserves more justification. Works of Bricaud, among others, suggest significant phytoplankton absorption there due to MAAs, particularly near the surface.
- The effect of S and T on water IOPs is very different in the bands investigated. S primarily contribute to backscattering but if a mean salinity is used, the change may be rather small. T affects primarily absorption in the NIR. Again, using an average T may be more than sufficient anywhere else.
- The final result that the best choice is bbp_tilde =0.002 and eta=0 deserves attention. It does not seem consistent with expectation from other studies.
- It is not clear why there seem to be no configuration with Fchl providing the vertical profile for a_phi, Fdom for a_cdom and b_bp for b_bp and NAP. If this one does not work well, please explain why you think it does not? What are the implications?
Dear authors: I am often wrong. If you feel my comments are ‘off mark’ please contact me and if convinced I will be happy to change them.
-
AC1: 'Reply on RC1', Elena Terzić, 09 Mar 2021
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2020-473/bg-2020-473-AC1-supplement.pdf
-
RC2: 'Comment on bg-2020-473', Anonymous Referee #2, 15 Feb 2021
This manuscripts presents a radiative transfer model that uses BGC-Argo float data to derive profiles of the underwater light field. This is a thorough, comprehensive study which tested different inherent optical property models/parameterizations, and validated them on float-measured downwelling irradiance and satellite observations.
Given the increasing use of BGC-Argo floats and autonomous gliders, methods to connect their biogeochemical measurements to optical properties/underwater light fields are important – this study is one of few attempting to fill that gap, and as such, is of interest to the wider community.
Ultimately, I think this manuscript could be a valuable contribution, however in its current form, I have some concerns/comments and areas that I think need a bit of clarification.
1) Some of the method details and corresponding results are a bit hard to follow. Sometimes it’s unclear what exact IOP parameterization is used where, and what parameterization the different results in the text are referring to. I think this could be simplified with a table or two summarizing these details.
2) Maybe I’m missing something in your methodology completely, but it is unclear to me why you test the individual OSM (optically significant material) constituent IOP models separately. You know that for your float data, you’re never in pure water or pure water + 1 OSM. You’re almost always going to be in your 6th IOP model group (i.e. all constituents). So you need to test your different individual OSM-IOP model parameterizations within this 6th IOP model group because it is the total IOPs that are important. For example, when testing your a_NAP model parameterizations (your 2nd IOP model group), you may get a better result (i.e. an irradiance that is closer to the measured value) for one of the parameterizations that is actually less correct. This is because that parameterization could be less correct in a way that is more correct in terms of the actual total IOP e.g. in this NAP case, perhaps your NAP model is incorrect, but it accurately captures CDOM, and CDOM is the dominant OSM in that sample, so overall, the total absorption is more correct.
Further comments are detailed below.
Specific comments
L7: Lefering et al (2020) used model, satellite and glider data for the type of three-platform comparison done here. [Development of bio-optical model for the Barents Sea to quantitatively link glider and satellite observations, Philosophical Transactions A: Mathematical, Physical and Engineering Sciences https://doi.org/10.1098/rsta.2019.0367]
Fig 1 & 2: units missing on the colour bar – number of floats? Number of days with profiles present over the 5 year period?
L75: you haven’t defined Kd yet
L75: The details on the Kd calculation are not clear. It seems like you are doing a depth averaged Kd rather than taking the derivative of the Ed profile (to get a profile of Kd). If so, what depth are you averaging over? Why do you need to have the first Ed measurement shallower than 1m?
L78: Are you saying that you tested all your different model parameterizations but then just removed the results that had an Ed difference of greater than 30%? What’s the justification for this?
L95, 99: missing backscattering coefficients
L142: What is the “this” at the start of the sentence “This is however present…” Can you please reword to make it clearer?
Eq 16 / L177: I don’t understand this. Why are you estimating aCDOM using Pope and Fry when you are using Mason et al elsewhere?
L178-179: in the text you have a_{ORIG}, but in Eq 16 you have a_w^{ORIG} – can you please make it consistent?
L195: No other OSMs affecting Kbio(380)? Can you expand on this in the text please.
Section 2.3.4: How do the min and max Chl values affect the results? You didn’t see any features/performance issues at the low and high ends of your data?
Eq 20: using chl as a bp model – what about NAP?
Are all IOP models derived for the “surface” and then extrapolated? What’s the “surface”?
Fig 6 & L310: what configurations are used for each of the OSMs here? The best aNAP and aCDOM results from Figs 4 and 5? You described a range of different scattering models – which one is shown in this figure?
L311-312: Suggest including that these decreases in RMSE and bias are in reference to the pure water simulation.
L323-326: what run is this paragraph referring to? I’m assuming the run with water + all OSMs. If so, what configurations are used for each of the OSMs? Can you please include those details?
Fig 7: The y-axis label isn’t defined anywhere
Fig 8 – 10: what is the depth bin or range for the model and float data shown in these figures?
I’m not sure I’m following the aNAP argument presented in section 3.2. You state the Kd comparisons improve when you remove the aNAP component – this makes sense because of your assumption that Kbio was only CDOM driven. Then you say you need the aNAP component to get the best Rrs retrievals, but offer no explanation. What’s going on here? It seems your Rrs model is inconsistent with your Kd model?
Technical corrections, typos etc
L10 “also incorporated” rather than “incorporated also”
L15 “also enabled the estimation of” rather than “enabled to estimate also”
L27 Suggest reordering, the second part of the sentence is fragmented. “To achieve this, and in order to follow up with the pace… in-situ platforms, the implementation…remains essential.”
L49-50: Same as above, suggest reordering “With their high vertical resolution profiles and high spatial coverage, BGC-Argo…”
L125: Haven’t defined PFT
L135 and a few other places: you have an indent after the equation where there shouldn’t be one.
L141 “of conveying” rather than “to convey”
L163 the Babin et al ref should be in parenthesis
L195-199: This sentence is slightly confusing. I think you’re saying the IOPs used for pure water are an upgrade because you use Mason et al and include a T-S correction. Suggest rewording along the lines of “Given the fact that the IOP models used here for pure water absorption are the new measurements of Mason et al, and a T-S correction is applied, …”
Eq 12 and 22: use bb’ and bb-tilde for backscattering ratio, please make it consistent throughout.
Fig 3: In the caption the formatting of the xlabels is different from the figure.
Fig 5 Caption: “twe” should be “the”
Citation: https://doi.org/10.5194/bg-2020-473-RC2 -
AC2: 'Reply on RC2', Elena Terzić, 09 Mar 2021
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2020-473/bg-2020-473-AC2-supplement.pdf
-
AC2: 'Reply on RC2', Elena Terzić, 09 Mar 2021
-
RC3: 'Comment on bg-2020-473', Zach Erickson, 16 Feb 2021
The authors use BGC-Argo data to compare IOPs (absorption and scattering) and AOPs (Kd and Rrs) estimated using simple water column radiative transfer, in situ measurements of water column constituents, and satellite remote sensing. The focus of this project is the Mediterranean, where the optically complex waters justify comparing different approaches. This paper directly compares different estimates of absorption spectra, and in many cases shows the utility both of more recent estimates of a_w, as well as the benefit of a regional PFT approach for pigment absorption.
While this research is definitely useful, in many places I struggled to understand what exactly the authors were doing. In particular, in Section 2.3 they list 6 different models that they test, comparing measured downwelling irradiance with that derived from different models following Eq. 1-12. Their first model is pure water absorption and scattering, for which they test three different estimates of a_w (Fig. 3). It wasn’t clear to me what downwelling irradiance observations they used to test the different a_w. From the text, it seems like they compared actual profiles of Ed to estimated profiles of Ed assuming clear water, but that doesn’t make any sense. Did they only use BGC-Argo profiles in very clear waters, and if so, how did they define that? Or did they have a “basic” version of the other IOPs that they used in Eq. 1-12, in which case, what was it, and how does it differ from experiment #6? I have a similar question for experiments 2-5; from the text it seems like the authors assume the only constituents are those specifically tested, but that doesn’t make sense unless you have measurement profiles for which that might be true. I’m sure I am missing something here.
Another point of confusion was how the authors use estimates of Chl from BGC-Argo. BGC-Argo floats do not measure Chl, they measure fluorescence from Chl (fChl). During the daytime, near the surface fChl can be reduced even though Chl is still elevated because of non-photochemical quenching (NPQ). There are a number of methods to correct for this, including specifically for BGC-Argo (e.g., Xing et al., 2018). Was any sort of NPQ correction applied to the fChl data? If not, I would be worried that the authors’ “Case I models” are tracking fChl and not Chl.
Finally, I was uncertain about how to interpret Figures 8-10 in light of the text on l. 254-256 and 332. Are the satellite values shown monthly climatological values? If so, I don’t see how these can be compared with Argo considering the time frames and spatial locations are different (that is, Argo doesn’t sample uniformly and because of cloud cover neither does the satellite, and they also have different temporal weights based on the different lifetimes of the satellites and Argo floats).
Figures in general: I recommend removing Figures 1 and the top panel of 2 (see comment below), and moving the information in Figures 3-6 into a single Table, after removing experiments that don’t make sense to include (see comments below). What would be more useful to see in a Figure would be some representative profiles of measured Chl, bbp, Ed, and modeled Ed. Instead of Figure 7, what would be more useful to see would be the regionality or where the measurements are close to or far away from the 1:1 line with the model.
Conclusions Section in general: A lot the text in the Conclusions does not logically follow from the results of this paper. For example, I don’t see a clear path between this paper and combining oxygen, nitrate, and pH in numerical models (l. 412-417), and I’m not convinced that this paper in particular demonstrates the need for inclusion of multi-spectral measurements (l. 423-424) or hyperspectral models (425-427). To be clear, I don’t necessarily disagree with any of these statements, I just don’t think these are conclusions or logical next steps that one would arrive at from reading this paper.
Smaller comments:
I don’t think it is useful to show how many BGC-Argo profiles didn’t have the right set of measurements at the right depths to do this analysis. If profiles were discarded because of data quality (rather than data availability) that could be useful to know – e.g., “a condition of less than 30% difference between modelled and computed Ed values was thus added which resulted in 147 profiles less” (l. 77-78), although I don’t quite understand what this means – but right now Figure 1 and the top panel of Figure 2 don’t add anything to the paper.
Sec. 2.2 could use more textual help for the equations – make sure to define variables in the same paragraph where they first appear and provide text to explain the equations. Also, when denoting variables associated with direct downward irradiance sometimes a subscript of “dir” is used (e.g., E_dir) and sometimes “d” (e.g., C_d); please pick one. Similarly for “dif” and “s”. Please also check in this section and throughout to make sure b and b_p mean scattering, and b_b and b_bp mean backscattering, and that variable dependencies are correct (e.g., a_w doesn’t depend on z).
l. 125: Define PFT.
The text around Eq. 13 is confusing; I don’t think this equation is actually helpful. The authors can just say in the text in Sections 2.3.3 and 2.3.4 that profiles either follow Chl, b_bp, or fDOM.
l. 141-142: I am confused by the statement that AOPs “to a certain extent remove the impact of the external environment’s variability”; by definition, AOPs are properties that depend on these external parameters.
Sec. 2.3: I found some of this section backwards – the text in l. 194-195 would be nice to have before the authors explain the text around Eq. 16. Also, considering the authors have chosen to use Mason et al. 2016 instead of Pope and Fry (1997), why show the test with the Pope and Fry-derived values at all? Similarly, in Fig. 5, why show the values without the T/S correction for a_w?
l. 229-230: If the estimated variability in this backscattering spectral power-law slope parameter is from 0-4, using a constant value of 2 seems overly simplistic. Some discussion about error and uncertainty as a result of this would be good to include.
l. 233: What final value of the backscattering ratio was used? What error and uncertainty is the result?
l. 336-337: The authors could be more explicit about what the two Kd estimates actually are here (both are float-derived). There should also be more discussion somewhere – why might IOPs be overestimated?
Need a statement about data availability somewhere (typically in Acknowledgements). Also, should specifically list which BGC-Argo floats were used.
Minor edits (not exhaustive):
l. 157-158: I think it is more correct to say that the absorption spectrum is often modeled (not “follows”) as an exponentially decreasing shape despite its heterogeneous biogeochemical composition.
l. 173: “expressed” should be “parameterized”? Similarly, in l. 225 “obtained” should be “estimated”.
l. 201-202: Organelli et al. (2017c) doesn’t provide spectra for wavelengths shorter than 400 nm; how do you get data at 380 nm? Also, how representative are these of Mediterannean waters? And finally, some explicit response should be given to that paper’s caveat that their results “were not intended for any algorithm development and/or validation” (to quote from their Conclusions).
l. 213-214: How many data points have Chl outside of these bounds? Would it not be preferable to just omit those profiles? If the authors keep them in, some discussion should be present about what error this introduces.
Eq. 21: negative sign missing in the exponent?
Fig. 5: Is it still considered Case I if the CDOM profile follows fDOM and not Chl?
l. 368: It should be noted that satellite-derived Rrs(412) is prone to large uncertainty, especially in optically complex waters; see e.g. Wei et al. (2020) and references therein.
Fig. 7: I don’t think the word BIOPTIMOD was anywhere in the main text?
Citation: https://doi.org/10.5194/bg-2020-473-RC3 -
AC3: 'Reply on RC3', Elena Terzić, 09 Mar 2021
The comment was uploaded in the form of a supplement: https://bg.copernicus.org/preprints/bg-2020-473/bg-2020-473-AC3-supplement.pdf
-
AC3: 'Reply on RC3', Elena Terzić, 09 Mar 2021
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 897 | 473 | 62 | 1,432 | 47 | 49 |
- HTML: 897
- PDF: 473
- XML: 62
- Total: 1,432
- BibTeX: 47
- EndNote: 49
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
