Accounting for surface reflectance variations along the satellite line of sight for UV-Vis ozone limb retrievals

Arosio, Carlo; Rozanov, Alexei; Rozanov, Vladimir; Orfanoz-Cheuquelaf, Andrea; Burrows, John P.; Malina, Edward; Boesch, Hartmut

doi:10.5194/egusphere-2026-5113

Preprints

https://doi.org/10.5194/egusphere-2026-5113

Preprints

04 Sep 2026

| 04 Sep 2026

Accounting for surface reflectance variations along the satellite line of sight for UV-Vis ozone limb retrievals

Carlo Arosio, Alexei Rozanov, Vladimir Rozanov, Andrea Orfanoz-Cheuquelaf, John P. Burrows, Edward Malina, and Hartmut Boesch

Abstract. Retrievals of stratospheric ozone profiles from space-borne limb-scatter measurements in the UV-Visible spectral range were found to produce a bias if the effective reflectance of the underlying scene has a strong gradient along the instrument line of sight. This effect is particularly pronounced when the underlying scene between the sub-satellite and the tangent point passes from desert to ocean or from cloudy to cloud-free conditions, especially over the ocean. For stratospheric ozone columns retrieved from Ozone Mapper and Profiler Suite Limb Profiler (OMPS-LP) observations, regional biases of up to 7–10 DU were detected. When applying the limb-nadir matching technique, i.e. combining stratospheric and total ozone column information to obtain the residual tropospheric column, these underlying scene related biases amount to about 10–30 % of typical tropospheric ozone column values.

In this manuscript we present an improved retrieval algorithm, which substantially mitigates the observed biases in the stratospheric and tropospheric ozone. The algorithm employs an extended version of the SCIATRAN radiative transfer model (v4.8) capable of accounting for variations of the effective reflectance of the underlying scene along the instrument line of sight. VIIRS observations are used to prescribe the variation of the effective reflectance of the underlying scene along the instrument line of sight in the radiative transfer model. The manuscript describes the implementation strategy, discusses the results in terms of stratospheric ozone changes with respect to the standard version and presents updated OMPS-LP/OMPS-NM and new OMPS-LP/TROPOMI tropospheric ozone column data sets along with the validation of the results with respect to correlative observations. Any retrieval of observations from solar limb sounding spectrometers is potentially affected by this issue and should take it into account.

Received: 26 Aug 2026 – Discussion started: 04 Sep 2026

Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.

Download & links

Preprint (PDF, 5418 KB)

Supplement (131 KB)

Download & links

Carlo Arosio, Alexei Rozanov, Vladimir Rozanov, Andrea Orfanoz-Cheuquelaf, John P. Burrows, Edward Malina, and Hartmut Boesch

Status: final response (author comments only)

RC1:
'Comment on egusphere-2026-5113', Anonymous Referee #1, 23 Sep 2026
The paper presents an improvement of the ozone retrieval method for satellite-based bright limb observations, using an updated version of SCIATRAN to account for surface reflectance gradients along the line of sight.
The improved method is applied to OMPS-LP observations and compared to a previous retrieval algorithm method which was not accounting for these gradients. Both retrievals are compared to other colocated satellite data (MLS, TROPOMI) and to HEGIFTOM ozone sonde data to assess the improvements of the improved retrieval in the stratosphere and in the troposphere.
The problem of the effect of strong reflectance gradients on bright limb observation profile retrievals is well explained and the paper proposes a significant improvement to the commonly used retrieval method, especially important for the UTLS region. The paper is relevant to the bright limb retrieval community. It is well organized and presented and I therefore recommend it for publication. Please find below my minor comments on the paper.
COMMENTS
Line 62: "This is especially the case for space borne instruments which register a full vertical image at once (such as OMPS-LP) because every tangent point of the limb image has somewhat different geographic location.": A figure to describe the geometry of the problem and the varying TP location would help to better understand the problem (either in the article or in the supplement).

Chappuis is mispelled "Chappius" in lines 109 and 172

Line 108 "For our study, we used surface reflectance values in the visible band at 673 nm, which we consider representative for the Chappius band, given that ocean reflectance and cloud albedo have a small wavelength dependence across the visible range.": A figure presenting the wavelength dependence of a typical ocean or cloud reflectance/albedo could be interesting (in the supplement) to support this claim.

Line 157: "The SCIATRAN version capable of accounting for variations of the surface albedo is referred to as v4.8 and the retrievals performed using this version are referred to in the following text as 2D-mode retrievals.": Does this SCIATRAN improvement add computational cost or is it negligible?

Line 204: "For each orbit, the location of the maximum of the underlying scene reflectance gradient (Δ) was taken as a reference.": I would add the location of Δ on Figure S2.

Line 249: "The positive bias, which is more evident in the Northern hemisphere, is related to a positive bias of the VIIRS surface albedo with respect to that retrieved from OMPS-LP, when using the standard 1D retrieval.": I find this sentence unclear. What is the link with the 1D retrieval? please clarify.

Line 251: "Furthermore, this discrepancy might also be caused by the uncertainty in the aerosol information used in ozone retrievals, which has the strongest impact in the Northern hemisphere.": Additional explanation is needed here, why is it more important in the northern hemisphere? Do you observe the same behavior in other years?

Line 297: "On top of these regional differences, that actually illustrate the effectiveness of the 2D-mode retrieval, we notice an overall negative bias in the southern middle and high latitudes.": do you have an explanation for this negative bias?

Line 304: "A climatology was used if the stratospheric profile bottom was above the tropopause.": Please elaborate; which climatology was used? Could this have an impact on the results? it is also unclear if this was only needed for TROPOMI-OMPS LNM or also for OMPS-LNM?

Comments on figures
Figure 1: why is this figure truncated at -25°/+25° in latitude? It could be interesting to compare the data presented here to figure 9 and figure 10 and have the same latitude range.

Figure 6: the right Y-axis ranging from 0 to 1 does not appear to correspond to anything and can be confusing with the surface reflectance.

Figure 7: it should be made more clear that this figure uses v5.6 retrievals in the legend. Could you also add another panel to present the same relative differences between OMPS-NPP v4.5 vs MLS and ozone sondes? it would be interesting to highlight the improvement of v5.6 and the altitude dependence of the correction.

Figure 8: Why does the 2D-mode show a slightly smaller bias in the southern mid-latitudes? I would expect a larger difference between the 1D and 2D modes at the equator, given the strong difference in TrOC obtained with the different SOCs (figure 9). Can you give some more explanations?

Figure 11: Y-axis should be TrOC [DU] instead of 'Troc DU'
Citation: https://doi.org/10.5194/egusphere-2026-5113-RC1
RC2: 'Comment on egusphere-2026-5113', Anonymous Referee #2, 05 Oct 2026

The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-5113/egusphere-2026-5113-RC2-supplement.pdf

Citation: https://doi.org/10.5194/egusphere-2026-5113-RC2
RC3:
'Comment on egusphere-2026-5113', Daniel Zawada, 05 Oct 2026
The authors present a method for accounting for surface reflectance variations in the along line of sight direction for limb scattering retrievals. The problem is motivated by an artifact observed in the current retrieval version, and the mitigation involves adapting the radiative transfer model used in the retrieval to account for surface albedo variations combined with auxiliary surface reflectance data from VIIRS. Overall the manuscript is well written, concise, and I recommend it be published following some minor revisions.
General Comments

——————————
I have seen these effects as well and have my own thoughts on why it happens but it has been a while since I have directly looked at it.
I agree that this bias is likely caused by a 3D varying scene underneath and that what has been done by the authors mitigates it. The exact mechanism for why it occurs I’m a little less sure about and isn’t discussed much in the manuscript. I don’t expect the authors to fully diagnose and explain what is happening, but some more commentary on possible mechanisms and why it happens should be added. I have a few general points below for the authors to think about.
As mentioned in the specific comments below, unless I am misunderstanding, the version of SCIATRAN is sort of pseudo-two dimensional, where every multiple scatter point assumes a 1D albedo. I don’t really know exactly what errors this introduces

I have always thought that this effect is not some complicated radiative transfer thing but actually just geometry. Higher tangent altitudes are geometry closer to the instrument and consequently see a different scene underneath. Essentially if there is a albedo gradient along the orbital track, you will get an effective albedo gradient as a function of tangent altitude for that image. This ends up being a problem in a 1d retrieval because usually you retrieve albedo at some higher altitude (~40 km) and then the lower altitudes that you actually care about see a different scene. It can also problems when you high altitude normalize where the 1d forward model does not account for the albedo gradient.

Related to the previous point, I am wondering if you have tried the 1D retrieval using the VIIRS reflectances (averaged over some effective region)? This would at least partially account for the geometry (if you used reflectances at the ~25km tangent point location)

I suppose overall what I am getting at is do you think it is possible this bias is entirely caused by geometry, the tangent point shifting closer to the instrument for higher lines of sight?
Specific Comments

——————————
l.16 “Any retrieval of observations from solar limb sounding spectrometers is potentially affected by this issue and should take it into account”: Perhaps too strong of a statement since (only?) OMPS-LP has the necessary secondary information (albedo from VIIRS) to apply the technique used here
l. 56 “As discussed by Oikarinen…”: this is tangent altitude dependent though correct?
l. 70 “and limb-scatter retrievals (Zawada et al. 2018)…”: we do in fact retrieve and use a 2d albedo in the retrieval for OMPS-LP. It was fairly crude in v1.3.0 but v2.1.0 has made some improvements here. So at least in theory we should be accounting for this, but there may be other issues. I am not sure if you have checked if we see the same artifacts with the new version or not, I am not suggesting you do the comparisons for the manuscript here though.
Section 2. Perhaps not here, but somewhere, you could mention that this can be extended to N21 OMPS-LP since N21 also contains VIIRS.
l 108. “We used surface reflectance values in the visible band at 673 nm...”: I am not too familiar with how VIIRS estimates reflectance, but is the VIIRS 673 nm channel influenced by the amount of ozone? I am also wondering if you have looked at what effect simply switching to the VIIRS albedo has in your 1d retrieval since it is at a different wavelength.
l. 148., and the following paragraph: If I understand correctly this isn’t really a truly 2d varying surface reflectance? Every multiple scatter calculation along the LOS is essentially a 1D calculation with a different albedo?
Figure 3, and the retrieval process: I would have actually expected that aerosol is significantly more influenced by albedo errors than the ozone field. Have you looked at the aerosol curtains, do they also suffer from the same artifact? It should be straightforward to also apply the same “2D” retrieval to the aerosol profiles, and this would couple back in to the ozone retrieval as well.
Figure 4: It would be useful here to see the % error, is the error only observed in the Chappuis retrieval range or does it extend higher?
l. 194 “We found generally good agreement between the three zonally averaged datasets with a small difference between OMPS-LP results and TROPOMI/VIIRS data”: It would be interesting to see if instead of using VIIRS you could do a 1D retrieval first, and then use that albedo dataset.
Section 6.1: I would suggest to show a few actual individual orbits, the 1d, 2d retrievals and their differences to MLS.
Section 6.3: For MLS it should be stated how it was converted to number density/altitude
l. 250 “A positive bias of the VIIRS surface albedo with respect to that retrieved from OMPS-LP”: Bias might not be the right word when they are at different wavelengths?
Figure 7: v4.5 should also be shown here
l. 332 “a bias can be introduced for all scenes below…”: All non-homogeneous scenes
l. 339 “the whole OMPS-LP time series (2012-2025)”: Up to 2025 is what you processed, but the whole time series goes into 2026?
l. 347: As mentioned earlier, OMPS-LP N21 should be mentioned here.
Citation: https://doi.org/10.5194/egusphere-2026-5113-RC3

Carlo Arosio, Alexei Rozanov, Vladimir Rozanov, Andrea Orfanoz-Cheuquelaf, John P. Burrows, Edward Malina, and Hartmut Boesch

Supplement

https://doi.org/10.5194/egusphere-2026-5113-supplement

Carlo Arosio, Alexei Rozanov, Vladimir Rozanov, Andrea Orfanoz-Cheuquelaf, John P. Burrows, Edward Malina, and Hartmut Boesch

Viewed

Total article views: 383 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
236	100	47	383	46	39	47

HTML: 236
PDF: 100
XML: 47
Total: 383
Supplement: 46
BibTeX: 39
EndNote: 47

Views and downloads (calculated since 04 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	183	70	31	284
Oct 2026	53	30	16	99

Cumulative views and downloads (calculated since 04 Sep 2026)

Month	HTML	PDF	XML	Total
Sep 2026	183	70	31	284
Oct 2026	53	30	16	99

Viewed (geographical distribution)

Total article views: 383 (including HTML, PDF, and XML) Thereof 383 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 14 Oct 2026

Short summary

We present an improved method for retrieving ozone profiles from OMPS satellite limb observations. The aim is to mitigate an artifact related to the variation of surface reflectivity along the satellite line of sight. We prescribe surface reflectivity from VIIRS to mitigate the artifact, particularly in regions with sharp changes in reflectivity, such as coastlines and transitions between clouds and ocean. Updated stratospheric ozone profiles and tropospheric ozone column products are presented.


Total:	0
HTML:	0
PDF:	0
XML:	0