the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 13 Jul 2026
Spectral correction factors for the removal of glint perturbations in above-water radiometry
Abstract. Spectral correction factors for the removal of glint perturbations in above-water radiometric measurements are theoretically computed for representative inland, coastal and open-ocean waters, accounting for spectral and atmospheric dependences. The simulation framework relies on the measurement protocol adopted by AERONET-OC and endorsed by the ocean color community to support the validation of satellite aquatic radiometric products. The theoretical computation of the correction factors, here termed glint correction factors or -factors, is performed in the 340–1020 nm spectral range by coupling i. a highly accurate plane-parallel scalar code to simulate the angular distribution of the spectral sky-radiance impinging at the water surface, with ii. a three-dimensional Monte Carlo code to model radiance reflections at a wind-roughened water surface. The impact of glint due to sky-radiance originating from the sun region (conventionally termed sun-glint) is separately discussed. Computed glint correction factors—for both the total sky radiance and its sole diffuse component—are available at Zenodo (https://doi.org/10.5281/zenodo.20609990).
- Preprint
(2166 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 19 Aug 2026)
-
RC1: 'Comment on essd-2026-473', Anonymous Referee #1, 15 Jul 2026
reply
-
AC1: 'Reply on RC1', Barbara Bulgarelli, 21 Jul 2026
reply
The comment was uploaded in the form of a supplement: https://essd.copernicus.org/preprints/essd-2026-473/essd-2026-473-AC1-supplement.pdf
-
AC1: 'Reply on RC1', Barbara Bulgarelli, 21 Jul 2026
reply
-
RC2: 'Comment on essd-2026-473', Anonymous Referee #2, 24 Jul 2026
reply
General appraisal
This manuscript describes a simulated data set of the reflection coefficients for sunlight interacting with a wind-roughened water surface. Coefficients are provided both for reflection of diffuse skylight only and for the sum of diffuse sky light and direct sun light. These coefficients are to be used in the processing of above-water radiometry measurements that aim at deriving the remote-sensing reflectance. During these measurements, one radiometer points towards the surface and therefore receives both photons that have interacted with the water body (the water-leaving radiance) and photons that have not entered the water and have only been reflected at the surface. This latter contribution is the one to be corrected for.
Therefore, the simulated data here presented should be quite useful for anyone attempting to process such above-water radiometry. The dependences that are here considered for these coefficients (wind speed, atmospheric conditions) have been previously assessed but probably not as thoroughly as in this work.
Overall, the paper provides all relevant information about the simulations and describes the results in great details. The models that were used are appropriate and have been run over a large range of conditions.
The number of Figures might be reduced a bit (18 Figures is a lot). Or some multi-panel Figures could be reduced to a single panel, for instance the one showing the largest variability, and a comment made in the text, like, for instance “results are similar for other conditions etc…”. For instance, Fig. 14 could be one panel only with 5 curves, one for each wind speed, showing the data for the 10° sun zenith angle and something like “.. results for other sun zenith angles show a lower dependence on the optical thickness”. Anyway, this is just a suggestion.
Although I understand this is a data paper, not a research paper, I think some comparison with existing values for r would be extremely useful. I suggest the Mobley 2015 data could be used for this (or maybe the authors have another draft in preparation where the impact of using these new coefficients is assessed?).
More fundamentally, I am not sure I understand the need to spend much time describing the simulations including the direct sun glint. My understanding of the processing of above-water radiometry is that data are first screened to eliminate all potential sun-glint “spots”. This is generally achieved by keeping the lowest few percent of the data set, which normally ensures that the only remaining contribution is the one by diffuse skylight. Section 3.2 alludes to this. Maybe another way to reduce the number of Figures?
The uncertainty analysis on the wind speed value (section 3.4) is relevant. I think it could be completed by a sensitivity analysis on the values of the 3 angles at play here (sun zenith, view and azimuth difference). Would be useful to know how much error can be tolerated on the latter, for instance (does not matter when measurements are taken from a fixed platform with a well-controlled geometry but does matter on a ship, for instance).
And, finally, a short section telling readers how to use the data in their processing could be relevant.
Few specific comments
- The description of the two components here considered that is given on lines 33-36 is clear, whereas I find the one in the abstract is a bit confusing. Therefore, I suggest the abstract is slightly modified in line with what is better said in the introduction.
- The use of the LR notation in Eq. (1) is not the wisest, because of the possible confusion with the radiance due to Rayleigh scattering.
- The ρD that appears line 130 should be better “justified”. Not sure I understand the logic. Or call it ρG (for glint); would be more explicit I guess.
- The fonts used for most Figures are really too small. All Figures should be revised to become way more legible. I also recommend that multi-panel Figures use letters to make easier the identification of the various panels.
- Fig. 3, right panel: I suggest stopping the red curve (aerosols) when there are no longer any aerosols in the atmosphere (a zero single scattering albedo looks weird).
- Fig. 6: I have a hard time understanding what the various panels are. This should be made much clearer ((c) to (h) are back for a sun zenith angle of 40° as in Fig. 5; is that right?)
Citation: https://doi.org/10.5194/essd-2026-473-RC2 -
RC3: 'Comment on essd-2026-473', Anonymous Referee #3, 24 Jul 2026
reply
Preliminary comment: I agree with the two previous reviews, which have already provided a good, constructive assessment.
I find the work by Bulgarelli and colleagues to be good and useful – it provides the underlying scientific and methodological context for a dataset (look-up table) of simulated sea surface reflectance factors, which are necessary for determining remote-sensing reflectance Rrs from above-water measurements – Rrs is the primary link between satellite remote sensing and biogeooptical properties of the water body. The factors are freely available and appear to be relatively easy to extract and apply. Therefore, I support the publication of this work.
However, unfortunately, the primary focus in designing the simulations for the ρ-data was placed on the reanalysis of AERONET-OC data. In my opinion, it would have been relatively straightforward to generate higher resolutions of the spectrum, the angular geometry and the wind speed (up to 15 m/s) through additional simulations. Could these higher resolutions be achieved without difficulty by interpolating, for example, the spectral ρ-factor for direct application to hyperspectral Rrs? There is a further dependence of ρ on aerosol types (maritime, mixed and continental), which can be roughly determined from in situ measurements. I view as critical the (presumably justified) dependence on the aerosol optical depth at 560 nm (tau560), an additional parameter that is usually not measured outside the AERONET network. These aspects significantly limit, or indeed prevent, the practical applicability of the ρ-factors to other networks such as WATERHYPERNET or typical in situ measurement campaigns. Furthermore, comparisons with other models and ρ-factors cannot be readily carried out, for example, to document any potential skylight-overcorrections. I see another problem in the restriction to just one azimuth angle (90°), which is certainly practical, but which rules out a great deal of flexibility in the field, particularly as best-practice protocols recommend using also a different angle – 135° – as the optimum for glint.
There are certainly models and initiatives that aim for simpler and broader applicability, but these are not mentioned here. These include the 3C model (Pitarch et al., 2020; Pitarch, 2026), which I consider to be very promising, and the ‘community processor’ HyperCP (e.g. Tilstone et al., 2025). I would suggest that you address the limitations and further developments regarding sea surface reflectance factors in the discussion.
Technical comment:
The figures are generally not very readable, certainly not in a printed version, and this is also a problem in the digital version where you can zoom in. The text size should be larger, and the ESSD Guidelines and general guidelines should be followed.
References:
Pitarch, J., Talone, M., Zibordi, G., & Groetsch, P. (2020). Determination of the remote-sensing reflectance from above-water measurements with the “3C model”: a further assessment. Optics Express, 28(11), 15885-15906.
Pitarch, J. (2026). A general model for sun and sky glint removal in above-water optical radiometry: mathematical description and Python code. Earth Science Informatics, 19(6), 78.
Tilstone, G. H., Jordan, T. M., Aurin, D., Białek, A., Deru, A., Ramsay, A., ... & Vendt, R. (2025). Radiometric field inter-comparison of fiducial reference measurements using an open source community processor. Optics Express, 33(7), 15756-15781.
Citation: https://doi.org/10.5194/essd-2026-473-RC3
Data sets
Spectral correction factors for the removal of glint perturbations in above-water radiometry B. Bulgarelli, D. D'Alimonte, G. Zibordi, and T. Kajyiama https://doi.org/10.5281/zenodo.20609990
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 150 | 30 | 12 | 192 | 10 | 11 |
- HTML: 150
- PDF: 30
- XML: 12
- Total: 192
- BibTeX: 10
- EndNote: 11
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
Spectral correction factors for the removal of glint perturbations in above-water radiometry
Review for ESSD
This paper presents updated estimates of the ρ look-up-table (LUT) for the glint removal from above-water measurements using Mobley’s approach, yet for an azimuth Δϕ=90°, which is not what recommended by Mobley, but the operational choice in the AERONET-OC network.
There are a number of facts from this paper that I find positive, and a number that I find negative. The editor may decide whether these latter ones compromise publication or not.
On the positive side I acknowledge the quality of the work. This is definitely the state of the art of radiative transfer, and both the finite element code and Monte Carlo simulator are unique tools that can provide the deepest insight on optics, maintaining the highest accuracy. The care on the atmospheric modelling is excellent, and hence, this whole scientific approach has high potential.
The motivation for a new ρ LUT beyond those that already exist is not clearly stated. I understand that the authors improved their modelling of the physical setup, but they also must explain what are the current physical or numerical inconsistencies that motivate the creation of yet another LUT. But even if the authors explained it, such a need is not very clear to me. There appears to be a proliferation of ρ’s in the recent years that is only adding entropy. I anticipate that, at the end, all but the small group of expert users will get confused and simply apply Mobley’s coefficients.
I downloaded the data. I cannot understand why the authors are releasing ρ values for Δϕ=90° only. I understand this is the azimuth of AERONET-OC and that the authors may not care much about other setups, but unfortunately that azimuth is only a value across a continuum of possible values. The authors are shooting themselves on the foot if they self-impose that limitation, given that the calculations were made for all, as it can be seen in the plots. Most users, especially those operating on smaller platforms, will choose Δϕ=135°, whereas automated sensors on ships face a different range of constraints, and have variable Δϕ. This definitely diminishes the impact of the paper.
The other big issue has to do with a basic requirement in science, which is reproducibility. The calculations come from proprietary software that cannot even be purchased. Hence, a reviewer or an independent scientist cannot reproduce the results. Yes, textual explanations on the simulation setup are provided, but that is largely insufficient. To me that would be a sufficient case for rejection, but it is the editor who has to make the decision after revising the journal policies.
On a smaller level of importance is coining the name “glint correction factor” for ρ. Now we don’t only have yet another set of ρ’s. We even have a new name. As a whole community, are we not even able to agree on a common nomenclature after 30 years that we are using these terms?
The comments on other authors’ results and on their own results are scientifically sound but excessively focused on the numerical point of view. I would prefer a more didactic tone, with more insight into physics, so that young researchers could use this document to learn about above-water radiometry. As an example, in lines 95-98, it is said that ρ tends to increase spectrally with increasing wind speed, driven by the spectral distribution of skylight. But why is that? Because the model is trying to explain reflected light with a significant contribution of sunlight using a reference measurement that is largely devoid of sunlight.
The authors claim on lines 84-85 that Mobley generated his ρ LUT using the Cox and Munk model without the 0.003 background term. I have checked the Hydrolight documentation, in particular the “Technical note 1” of Hydrolight 5, and the 0.003 term indeed appears. Authors shall clarify this.
The labels of Figures 5, 6, 7, 8 are hard to read due to the font size, and Figures 9, 10, 11, 12, 13, 14, 15, 16, 17 and 18 are totally unreadable. In addition, there appear to be rendering issues, so the resolution is poor.
In some cases, where multiple panels are displayed on the same figure, the axis scale is kept fixed, at the cost of being unable to grasp the numbers that are actually displayed. Example: Figure 9.
The abstract must synthetise the paper as a whole. So far it lacks a summary on the numerical results and a comparison with former ρ estimates.
I am missing information on the simulation of the optical proper of the water column. I understand that this aspect is of minor importance, given that the subject under examination is the reflection of the surface, but one still needs to make a simulation using specific water conditions. The authors are encouraged to disclose this.
The introduction states that the water leaving radiance is the primary product of aquatic remote sensing. This statement is not correct. The primary product is reflectance, as it is the quantity provided by satellite products, and the source of information that is extracted by remote sensing algorithms.
I cannot understand the distinction between L_i and L_sky in equations 1 and 2. I believe they are the same. In fact, if one inspects analog equations by other authors such as Mobley and Harmel, one finds that they use the same variable for both. Hence, authors must unify the nomenclature. Indeed, equation 3 confirms that L_i is nothing more than L_sky for a fixed geometry. That does not justify a new variable name.
Related to the previous comment, a table of mathematical symbols would highly help the reading.
Authors may revise the prose with the aid of a native speaker or a LLM. For instance, I find here several examples of what I call “JRC English”, namely expressions such as “It is highlighted that…”, “Still recognizing that…”, “Acknowledging that…”. Also, the division of the text in blocks might be improved. In particular, I find the Summary and Conclusions section rather disorganized.
I hope this set of comments is of help.