An Assessment of Lunar Photometry in AERONET

Schafer, Joel; Eck, Thomas F.; Slutsker, Ilya; Gupta, Pawan; Sorokin, Mikhail; Sinyuk, Alexander; Smirnov, Alexander

doi:10.5194/egusphere-2026-1506

Preprints

https://doi.org/10.5194/egusphere-2026-1506

Preprints

26 Mar 2026

| 26 Mar 2026

An Assessment of Lunar Photometry in AERONET

Joel Schafer, Thomas F. Eck, Ilya Slutsker, Pawan Gupta, Mikhail Sorokin, Alexander Sinyuk, and Alexander Smirnov

Abstract. AERONET has been acquiring direct beam lunar observations at six wavelengths from 440 to 1640 nm from most newer model T CIMEL sun-sky radiometers in the network for many years and producing a night-time AOD data set, which currently includes observations at 492 sites dating back as far as 2014. The new dataset of lunar AOD now uses an updated empirical correction for the ROLO lunar irradiance model and has been extensively analyzed for all long-term lunar-capable sites by evaluating the continuity of AOD between solar and lunar measurements during limited temporal windows. Comparison of daytime and nighttime AOD measurements shows good agreement at all sites with more than a decade of data with mean differences typically within ±0.01 for AOD_{440 nm} and similar or better agreement at longer wavelengths. Comparisons during 3-hour transition periods between day and night observations during conditions of low and stable AOD found statistically negligible differences of AOD at all common wavelengths as well as for Ångström Exponent, column water vapor and AOD_fine and AOD_coarse parameters from SDA. The lunar AOD product demonstrates consistency with solar AOD across diverse aerosol conditions at AERONET sites globally, validating the empirical correction approach.

Received: 17 Mar 2026 – Discussion started: 26 Mar 2026

Competing interests: At least one of the (co-)authors is a member of the editorial board of Atmospheric Measurement Techniques.

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, 5608 KB)

Notice on discussion status
The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.
Preprint (5608 KB)

Supplement (229 KB)

Download & links

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Journal article(s) based on this preprint

15 Oct 2026

An assessment of lunar photometry in AERONET

Joel Schafer, Thomas F. Eck, Ilya Slutsker, Pawan Gupta, Mikhail Sorokin, Alexander Sinyuk, and Alexander Smirnov

Atmos. Meas. Tech., 19, 6487–6502, https://doi.org/10.5194/amt-19-6487-2026,https://doi.org/10.5194/amt-19-6487-2026, 2026

Short summary

Joel Schafer, Thomas F. Eck, Ilya Slutsker, Pawan Gupta, Mikhail Sorokin, Alexander Sinyuk, and Alexander Smirnov

Interactive discussion

Status: closed

CC1:
'Comment on egusphere-2026-1506', Matthijs Krijger, 26 Apr 2026

The authors provide a nice detailed description of all their calibration steps, including a novel LUT correction to the well-established ROLO model that seems to work well. In figure 1 the impact of the ROLO correction for 440, 500, 675, 870, 1020, 1640nm are shown, but in figure 2 the LUT corrections are only shown for 440nm and 675nm. Without knowledge of the correction at other wavelenghts it is impossible for other scientists to investigate whether or not the corrections are physical/reasonable and/or replicate the results.
As such please provide also the ROLO corrections for the other wavelenghts., either visually and/or preferable an open dataset for application by other users.

Citation: https://doi.org/10.5194/egusphere-2026-1506-CC1
- AC1: 'Reply on CC1', Joel Schafer, 30 Apr 2026
  
  This is a valuable suggestion. A table of the derived ROLO correction coefficients for each CIMEL wavelength will be included as supplemental data. Thanks for the comment.
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC1
CC2:
'Comment on egusphere-2026-1506', Hugh Kieffer, 04 May 2026

Leading digits in comments are line number in the preprint
This is a significant paper and the large effort described will aid in approaching absolute lunar calibration.

------------------ Major -----------------

As you state, it is widely thought that ROLO model has biases. Better models have been available for several years.

You may find the following comprehensive review helpful.

"Activities to Promote the Moon as an Absolute Calibration Reference"

Z. Jing et al, Remote Sens. 2023, 15, 2431. https://doi.org/10.3390/rs15092431
58: Use of the LIME model for this work would be problematic, as they are based upon the same thru-the-atmosphere methodology and share data observations. This work and LIME would best be derived in parallel! although there are no authors in common.

54+. There is a newer lunar spectral irradiance model that would be useful.

H. H. Kieffer, ''Multiple-instrument-based spectral irradiance of the Moon'',

J. Applied Remote Sensing, Vol. 16(3), 038502 (2022)

https://doi.org/10.1117/1.JRS.16.038502 (open access)

As listed in Table 7 of that SLIM (now SLIMM) paper, ROLO is 3 to 8% below SLIMM.
Over 400 to 1040 nm, the SLIMM model is within 1% of the carefully calibrated air-LUSI results (discussed within GSICS, but not yet published)
Fig 1. These values look close to what the SLIMM model would forecast.
~75 The variation of lunar irradiance with sub-viewer lunar coordinates ("libration") is about 1% for modest phase and more at large phase. Libration should be considered in a correction LUT. 600 observations should be enough to assess the magnitude of libration from your data alone. Or you could use the model based on lunar maps (e.g., SLIMM Table 3). Hopefully, this will reduce the scatter in Figure 2.
Section 3.1 Good discussion of lunar irradiance properties.
Fig 5 and 6. Good figures!
Section 5: The near-in-time solar and lunar observations by the same instruments, especially at the two high-elevation sites, are key to this paper.

All things that might affect these need thorough and prominent quantitative discussion. E.g. ,instrument non-linearity, exposure timing resolution and accuracy, sensitivity to sky radiance level, ADC resolution at low light levels, ... . The sensitivity to AOD is already well discussed.
Section 5.2 Overall bias message is clear, but the concept behind 440nm exclusion and Fig. 15 are hard to follow.
245. Say where is Palangkaraya and its elevation.
Fig 15a. Move both interior labels down so that 'Solar' does not look like a subtitle to the upper part

252:253 Meaning unclear. Is 'range ' a verb or noun here?

Citation: https://doi.org/10.5194/egusphere-2026-1506-CC2
- AC2: 'Reply on CC2', Joel Schafer, 29 Jul 2026
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1506/egusphere-2026-1506-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC2
RC1:
'Comment on egusphere-2026-1506', Anonymous Referee #1, 13 May 2026

This manuscript presents an extensive assessment of AERONET nighttime aerosol optical depth (AOD) retrievals based on lunar photometry and empirical correction of the ROLO lunar irradiance model. The work is timely and important because nighttime AOD observations are increasingly valuable for aerosol monitoring, air quality applications, wildfire smoke tracking, and studies of diurnal aerosol variability. The manuscript is generally well organized and contains a large amount of useful analysis based on long-term global AERONET observations. The effort to evaluate continuity between solar and lunar retrievals across hundreds of sites is particularly valuable.
Major comments:
The current validation strategy relies heavily on continuity between solar and lunar AOD during day–night transition periods under low and stable aerosol conditions (AOD440 ≤ 0.1 with low variability). While this is a reasonable approach for isolating calibration consistency and minimizing atmospheric variability, it may introduce a selection bias toward relatively homogeneous and low-loading atmospheric states. Consequently, the presented analysis primarily demonstrates agreement under favorable retrieval conditions rather than across the broader range of nighttime aerosol environments that are often of greatest scientific and operational interest.

Although the empirical correction is applied at the irradiance calibration level and is therefore not explicitly AOD-dependent, it is unclear whether retrieval performance remains equivalent under high-AOD conditions where signal levels are substantially reduced and atmospheric variability is often larger. Additional discussion of how calibration uncertainty, low-signal effects, and retrieval bias may vary with aerosol loading would strengthen the manuscript.

In addition, the manuscript later notes that high-AOD lunar observations are more likely to be excluded because of low-signal limitations, particularly at shorter wavelengths. This raises the possibility of an additional sampling bias in the validation dataset.

It would strengthen the manuscript if the authors discussed more explicitly how these filtering criteria may limit the representativeness of the validation framework, particularly for high-AOD, polluted, smoke-affected, or rapidly evolving nighttime conditions. Relatedly, would the reported agreement statistics change materially under elevated AOD conditions?
Minor:
Line 43: Vo should be V₀

Figure 2: The legend titles should be made consistent (e.g., use either “channel” or “wavelength” for both figures). Additionally, why do the phase-angle dependence patterns differ between the two wavelengths?

Section 4: The manuscript would benefit from clearer specification of the data quality levels used throughout each analysis section. While Section 5 explicitly references Level 2 (L2) AOD data, it is less clear whether the transition analyses in Section 4 also rely exclusively on L2 observations. It would be helpful to clarify whether “L2” solar and lunar products should be interpreted as equivalent in quality assurance rigor and discuss any implications for the validation results.

Line 137: The statement that deviations “diminish with increasing wavelength” would be stronger if quantitative wavelength-dependent results or statistics were provided.

Line146: Figures 7 and 8 would benefit from additional discussion and interpretation. In particular, the spread of the day–night AOD differences appears to increase substantially at higher AOD values. This may reflect more rapidly evolving aerosol conditions during transition periods, but it could also indicate increasing retrieval uncertainty or residual bias under high-loading conditions where lunar signal limitations become more important. Please discuss the likely causes of the increased scatter and whether retrieval performance degrades systematically at elevated AOD.

Line 252: Please provide a quantitative description for “the great majority.” More generally, several statements throughout the manuscript rely on qualitative descriptors such as “good agreement,” “minor,” “very small,” “not notably dependent,” and “the great majority” without quantitative definition. I recommend providing explicit numerical thresholds, percentages, or statistical metrics wherever possible to improve interpretability and reproducibility.

Some figure labels and color-bar annotations are difficult to read because of small font sizes (e.g., Figures 7 and 8).

Citation: https://doi.org/10.5194/egusphere-2026-1506-RC1
- AC3: 'Reply on RC1', Joel Schafer, 29 Jul 2026
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1506/egusphere-2026-1506-AC3-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC3
RC2:
'Comment on egusphere-2026-1506', Anonymous Referee #2, 11 Jun 2026

The manuscript makes a thorough assessment of the lunar-based aerosol optical depth within AERONET. The performance of this new AERONET product is successful as shown by a robust statistical analysis and comparison to the well-established daytime AOD, including cloud screening. The explanations on the limitations of the technique are very illustrative. The paper is worth publication in AMT upon correction of several aspects that are not sufficiently explained (minor revision):
1. The uncertainty of the lunar-based AOD is not provided. For daytime, the 0.01-0.02 of Holben et al 1998 paper is widely known. What is it for AERONET lunar AOD product? Should it depend on airmass, lunar phase angle, AOD itself…? Uncertainty assessment is crucial if you intend to provide an assessment of this product.
Nothing is said on the quality assurance of lunar-based data. Will this product have a level 2? Are the quality criteria the same as for solar version 3 AOD?
2. The method for correction of the ROLO needs a much more detailed description. As far as I know, it is the first AERONET publication on lunar AOD. You need to explain it, from the beginning. Give formulation, details and references on the applied calibration (lunar Langley plot, etc.). If all calculation of lunar AOD is just the same as for solar AOD in version 3, please state it and refer to Giles et al paper. Even astronomical libraries (SPICE?) need to be reported and referenced.

For reproducibility, the correction factors for each wavelength must be provided, together with descriptive statistics, probably binned for phase angle intervals. For instance, the noisy data in Figure 2 are worth some comment (reasons?) and quantification.
3. The references to prior work in the field are rather incomplete and, in some cases, also inaccurate. I miss citation to prior work in lunar photometry within AERONET (e.g. Berkoff et al 2011, several Barreto et al papers), to studies reporting ROLO inaccuracies (e.g. Lacherade et al 2013 in comparison to satellite, or the assessment by Stone et al 2020, Kieffer 2022). The correction by Roman et al is not based on star photometry; it relies on day-night continuity in AOD. Star photometry is used as an independent measurement to validate the results. The NASA Air Lusi project or other lunar models are not mentioned either.
Other comments:
- Line 24: there must be much more targeted references about sun photometer observations in Ny Alesund (and polar regions in general).
- Line 26-28: if you speak about the AERONET lunar method, this is the first time it is described, so you cannot mention it as if it were well known.
- Line 28-35: the two papers cited here are just examples of usage of the preliminary (and undescribed) lunar AERONET product. Are these the only 2 papers using lunar AOD until now? Or are they just examples? Why these?
- Line 36-39: the final goal of the paper is the assessment of the product, but the description of the retrieval method is key for transparency and must be part of the paper. So far all AERONET products have been extensively described in key publications (Holben, Dubovik, Eck, Smirnov, Giles, Sinyuk, O’Neill…).
- Figure 3: visual inspection with such AOD levels is not very indicative of the performance. Try Mauna Loa. On the other hand, the examples shown here illustrate that, for most of the sites, you don’t have any practical difference, which is also important and an indirect indication of lunar-based AOD uncertainty.
- If the performance of the night-time AOD is based on the comparison with daytime values, then the possible benefit of finding day-night cycles or systematic differences (lines 15-20) is somehow contradictory. Please clarify.
- Line 93: ‘mechanical’ could maybe be replaced with ‘geometrical’?
- Section 3.1: ‘libration’ is not explicitly mentioned. Please include it as it is a concept that is pertinent to lunar irradiance modeling.
- Section 3.1.1: solar ‘declination’ is not mentioned. More importantly, the discussion really lacks an example for high latitudes (e.g. the Ny Alesund site previously mentioned by the authors). Figure 4 could be enriched with high, low and mid-latitude cases. Nighttime photometry is key in polar regions. The number of possible lunar measurements can range from 0 to 24h within a day depending on the Sun elevation.
- Line 130: if the day-night differences depend on calibration, they should be insensitive to AOD. Why limiting to low AOD cases? Isn’t aerosol variability an issue in this comparison exercise?
- Lines 138-140: the sentence is unclear to me. Please rephrase. Where is this ‘point of reference’ shown? What do you mean with: ‘with a similar time window gap entirely within the solar day and lunar night’ ?
- Figures 5, 6, 7: please state in the caption what the color coding means.
- Figure 7: the rmse is very small (partly by the huge number of data). But cases of large AOD differences (>0.1) are found in the plots, even for low AOD cases. How can this happen? Is it related to lunar phase? Table 1 indicates STD is reduced for longer wavelengths. Is there some explanation for that?
- Line 147: ‘excluding’ should probably read ‘excluded’.
- Table 2 caption looks somehow incomplete.
- Line 159: please give quantification of global % of data removed by the curvature threshold. If it’s marginal (as it seems from the discussion), it shouldn’t be an issue for the network.
- Line 189-190: ‘sunphotometer’ may be replaced by ‘photometer’. More importantly, the Vo/1500 criterion sounds a bit arbitrary and could be better justified (or referenced). And once this is applied in the AOD retrieval scheme, of course the limit on AOD is clear and you don’t actually need any real data to justify it. The Beer law and the airmass give it straight. The red curve in Fig.12 can be a simple calculation, it doesn’t need data to be supported.
This is maybe not so obvious for the lunar data, especially because the retrieval scheme is not described. Some example of Sun and Moon Vo’s (typical values for a Cimel T) would help to understand the differences, as well as the noise levels. Ultimately the signal-to-noise ratio appears to be more limiting factor in the lunar data than the diffuse light (sky) contribution within the instrument field of view, which is in turn the limiting factor with the Sun (wavelength differences may occur, though).
- Line 267: why not showing 870 or 1020 nm here if those wavelengths are more pertinent to cloud screening?
- Line 274-275: this is not a conclusion of this study. It’s rather a general comment that could be moved to the introduction.
- Line 282: the usable hours of Sun and Moon strongly depend on latitude and season, see comment on section 3.1.1. This part of the conclusions can be enhanced with a global vision (instead of a mid-latitude vision only).

Citation: https://doi.org/10.5194/egusphere-2026-1506-RC2
- AC4: 'Reply on RC2', Joel Schafer, 29 Jul 2026
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1506/egusphere-2026-1506-AC4-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC4

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Joel Schafer on behalf of the Authors (04 Aug 2026) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (06 Aug 2026) by Daniel Perez-Ramirez

RR by Anonymous Referee #3 (20 Aug 2026)

RR by Anonymous Referee #1 (20 Aug 2026)

ED: Publish as is (15 Sep 2026) by Daniel Perez-Ramirez

AR by Joel Schafer on behalf of the Authors (23 Sep 2026) Author's response Manuscript

Interactive discussion

Status: closed

CC1:
'Comment on egusphere-2026-1506', Matthijs Krijger, 26 Apr 2026

The authors provide a nice detailed description of all their calibration steps, including a novel LUT correction to the well-established ROLO model that seems to work well. In figure 1 the impact of the ROLO correction for 440, 500, 675, 870, 1020, 1640nm are shown, but in figure 2 the LUT corrections are only shown for 440nm and 675nm. Without knowledge of the correction at other wavelenghts it is impossible for other scientists to investigate whether or not the corrections are physical/reasonable and/or replicate the results.
As such please provide also the ROLO corrections for the other wavelenghts., either visually and/or preferable an open dataset for application by other users.

Citation: https://doi.org/10.5194/egusphere-2026-1506-CC1
- AC1: 'Reply on CC1', Joel Schafer, 30 Apr 2026
  
  This is a valuable suggestion. A table of the derived ROLO correction coefficients for each CIMEL wavelength will be included as supplemental data. Thanks for the comment.
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC1
CC2:
'Comment on egusphere-2026-1506', Hugh Kieffer, 04 May 2026

Leading digits in comments are line number in the preprint
This is a significant paper and the large effort described will aid in approaching absolute lunar calibration.

------------------ Major -----------------

As you state, it is widely thought that ROLO model has biases. Better models have been available for several years.

You may find the following comprehensive review helpful.

"Activities to Promote the Moon as an Absolute Calibration Reference"

Z. Jing et al, Remote Sens. 2023, 15, 2431. https://doi.org/10.3390/rs15092431
58: Use of the LIME model for this work would be problematic, as they are based upon the same thru-the-atmosphere methodology and share data observations. This work and LIME would best be derived in parallel! although there are no authors in common.

54+. There is a newer lunar spectral irradiance model that would be useful.

H. H. Kieffer, ''Multiple-instrument-based spectral irradiance of the Moon'',

J. Applied Remote Sensing, Vol. 16(3), 038502 (2022)

https://doi.org/10.1117/1.JRS.16.038502 (open access)

As listed in Table 7 of that SLIM (now SLIMM) paper, ROLO is 3 to 8% below SLIMM.
Over 400 to 1040 nm, the SLIMM model is within 1% of the carefully calibrated air-LUSI results (discussed within GSICS, but not yet published)
Fig 1. These values look close to what the SLIMM model would forecast.
~75 The variation of lunar irradiance with sub-viewer lunar coordinates ("libration") is about 1% for modest phase and more at large phase. Libration should be considered in a correction LUT. 600 observations should be enough to assess the magnitude of libration from your data alone. Or you could use the model based on lunar maps (e.g., SLIMM Table 3). Hopefully, this will reduce the scatter in Figure 2.
Section 3.1 Good discussion of lunar irradiance properties.
Fig 5 and 6. Good figures!
Section 5: The near-in-time solar and lunar observations by the same instruments, especially at the two high-elevation sites, are key to this paper.

All things that might affect these need thorough and prominent quantitative discussion. E.g. ,instrument non-linearity, exposure timing resolution and accuracy, sensitivity to sky radiance level, ADC resolution at low light levels, ... . The sensitivity to AOD is already well discussed.
Section 5.2 Overall bias message is clear, but the concept behind 440nm exclusion and Fig. 15 are hard to follow.
245. Say where is Palangkaraya and its elevation.
Fig 15a. Move both interior labels down so that 'Solar' does not look like a subtitle to the upper part

252:253 Meaning unclear. Is 'range ' a verb or noun here?

Citation: https://doi.org/10.5194/egusphere-2026-1506-CC2
- AC2: 'Reply on CC2', Joel Schafer, 29 Jul 2026
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1506/egusphere-2026-1506-AC2-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC2
RC1:
'Comment on egusphere-2026-1506', Anonymous Referee #1, 13 May 2026

This manuscript presents an extensive assessment of AERONET nighttime aerosol optical depth (AOD) retrievals based on lunar photometry and empirical correction of the ROLO lunar irradiance model. The work is timely and important because nighttime AOD observations are increasingly valuable for aerosol monitoring, air quality applications, wildfire smoke tracking, and studies of diurnal aerosol variability. The manuscript is generally well organized and contains a large amount of useful analysis based on long-term global AERONET observations. The effort to evaluate continuity between solar and lunar retrievals across hundreds of sites is particularly valuable.
Major comments:
The current validation strategy relies heavily on continuity between solar and lunar AOD during day–night transition periods under low and stable aerosol conditions (AOD440 ≤ 0.1 with low variability). While this is a reasonable approach for isolating calibration consistency and minimizing atmospheric variability, it may introduce a selection bias toward relatively homogeneous and low-loading atmospheric states. Consequently, the presented analysis primarily demonstrates agreement under favorable retrieval conditions rather than across the broader range of nighttime aerosol environments that are often of greatest scientific and operational interest.

Although the empirical correction is applied at the irradiance calibration level and is therefore not explicitly AOD-dependent, it is unclear whether retrieval performance remains equivalent under high-AOD conditions where signal levels are substantially reduced and atmospheric variability is often larger. Additional discussion of how calibration uncertainty, low-signal effects, and retrieval bias may vary with aerosol loading would strengthen the manuscript.

In addition, the manuscript later notes that high-AOD lunar observations are more likely to be excluded because of low-signal limitations, particularly at shorter wavelengths. This raises the possibility of an additional sampling bias in the validation dataset.

It would strengthen the manuscript if the authors discussed more explicitly how these filtering criteria may limit the representativeness of the validation framework, particularly for high-AOD, polluted, smoke-affected, or rapidly evolving nighttime conditions. Relatedly, would the reported agreement statistics change materially under elevated AOD conditions?
Minor:
Line 43: Vo should be V₀

Figure 2: The legend titles should be made consistent (e.g., use either “channel” or “wavelength” for both figures). Additionally, why do the phase-angle dependence patterns differ between the two wavelengths?

Section 4: The manuscript would benefit from clearer specification of the data quality levels used throughout each analysis section. While Section 5 explicitly references Level 2 (L2) AOD data, it is less clear whether the transition analyses in Section 4 also rely exclusively on L2 observations. It would be helpful to clarify whether “L2” solar and lunar products should be interpreted as equivalent in quality assurance rigor and discuss any implications for the validation results.

Line 137: The statement that deviations “diminish with increasing wavelength” would be stronger if quantitative wavelength-dependent results or statistics were provided.

Line146: Figures 7 and 8 would benefit from additional discussion and interpretation. In particular, the spread of the day–night AOD differences appears to increase substantially at higher AOD values. This may reflect more rapidly evolving aerosol conditions during transition periods, but it could also indicate increasing retrieval uncertainty or residual bias under high-loading conditions where lunar signal limitations become more important. Please discuss the likely causes of the increased scatter and whether retrieval performance degrades systematically at elevated AOD.

Line 252: Please provide a quantitative description for “the great majority.” More generally, several statements throughout the manuscript rely on qualitative descriptors such as “good agreement,” “minor,” “very small,” “not notably dependent,” and “the great majority” without quantitative definition. I recommend providing explicit numerical thresholds, percentages, or statistical metrics wherever possible to improve interpretability and reproducibility.

Some figure labels and color-bar annotations are difficult to read because of small font sizes (e.g., Figures 7 and 8).

Citation: https://doi.org/10.5194/egusphere-2026-1506-RC1
- AC3: 'Reply on RC1', Joel Schafer, 29 Jul 2026
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1506/egusphere-2026-1506-AC3-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC3
RC2:
'Comment on egusphere-2026-1506', Anonymous Referee #2, 11 Jun 2026

The manuscript makes a thorough assessment of the lunar-based aerosol optical depth within AERONET. The performance of this new AERONET product is successful as shown by a robust statistical analysis and comparison to the well-established daytime AOD, including cloud screening. The explanations on the limitations of the technique are very illustrative. The paper is worth publication in AMT upon correction of several aspects that are not sufficiently explained (minor revision):
1. The uncertainty of the lunar-based AOD is not provided. For daytime, the 0.01-0.02 of Holben et al 1998 paper is widely known. What is it for AERONET lunar AOD product? Should it depend on airmass, lunar phase angle, AOD itself…? Uncertainty assessment is crucial if you intend to provide an assessment of this product.
Nothing is said on the quality assurance of lunar-based data. Will this product have a level 2? Are the quality criteria the same as for solar version 3 AOD?
2. The method for correction of the ROLO needs a much more detailed description. As far as I know, it is the first AERONET publication on lunar AOD. You need to explain it, from the beginning. Give formulation, details and references on the applied calibration (lunar Langley plot, etc.). If all calculation of lunar AOD is just the same as for solar AOD in version 3, please state it and refer to Giles et al paper. Even astronomical libraries (SPICE?) need to be reported and referenced.

For reproducibility, the correction factors for each wavelength must be provided, together with descriptive statistics, probably binned for phase angle intervals. For instance, the noisy data in Figure 2 are worth some comment (reasons?) and quantification.
3. The references to prior work in the field are rather incomplete and, in some cases, also inaccurate. I miss citation to prior work in lunar photometry within AERONET (e.g. Berkoff et al 2011, several Barreto et al papers), to studies reporting ROLO inaccuracies (e.g. Lacherade et al 2013 in comparison to satellite, or the assessment by Stone et al 2020, Kieffer 2022). The correction by Roman et al is not based on star photometry; it relies on day-night continuity in AOD. Star photometry is used as an independent measurement to validate the results. The NASA Air Lusi project or other lunar models are not mentioned either.
Other comments:
- Line 24: there must be much more targeted references about sun photometer observations in Ny Alesund (and polar regions in general).
- Line 26-28: if you speak about the AERONET lunar method, this is the first time it is described, so you cannot mention it as if it were well known.
- Line 28-35: the two papers cited here are just examples of usage of the preliminary (and undescribed) lunar AERONET product. Are these the only 2 papers using lunar AOD until now? Or are they just examples? Why these?
- Line 36-39: the final goal of the paper is the assessment of the product, but the description of the retrieval method is key for transparency and must be part of the paper. So far all AERONET products have been extensively described in key publications (Holben, Dubovik, Eck, Smirnov, Giles, Sinyuk, O’Neill…).
- Figure 3: visual inspection with such AOD levels is not very indicative of the performance. Try Mauna Loa. On the other hand, the examples shown here illustrate that, for most of the sites, you don’t have any practical difference, which is also important and an indirect indication of lunar-based AOD uncertainty.
- If the performance of the night-time AOD is based on the comparison with daytime values, then the possible benefit of finding day-night cycles or systematic differences (lines 15-20) is somehow contradictory. Please clarify.
- Line 93: ‘mechanical’ could maybe be replaced with ‘geometrical’?
- Section 3.1: ‘libration’ is not explicitly mentioned. Please include it as it is a concept that is pertinent to lunar irradiance modeling.
- Section 3.1.1: solar ‘declination’ is not mentioned. More importantly, the discussion really lacks an example for high latitudes (e.g. the Ny Alesund site previously mentioned by the authors). Figure 4 could be enriched with high, low and mid-latitude cases. Nighttime photometry is key in polar regions. The number of possible lunar measurements can range from 0 to 24h within a day depending on the Sun elevation.
- Line 130: if the day-night differences depend on calibration, they should be insensitive to AOD. Why limiting to low AOD cases? Isn’t aerosol variability an issue in this comparison exercise?
- Lines 138-140: the sentence is unclear to me. Please rephrase. Where is this ‘point of reference’ shown? What do you mean with: ‘with a similar time window gap entirely within the solar day and lunar night’ ?
- Figures 5, 6, 7: please state in the caption what the color coding means.
- Figure 7: the rmse is very small (partly by the huge number of data). But cases of large AOD differences (>0.1) are found in the plots, even for low AOD cases. How can this happen? Is it related to lunar phase? Table 1 indicates STD is reduced for longer wavelengths. Is there some explanation for that?
- Line 147: ‘excluding’ should probably read ‘excluded’.
- Table 2 caption looks somehow incomplete.
- Line 159: please give quantification of global % of data removed by the curvature threshold. If it’s marginal (as it seems from the discussion), it shouldn’t be an issue for the network.
- Line 189-190: ‘sunphotometer’ may be replaced by ‘photometer’. More importantly, the Vo/1500 criterion sounds a bit arbitrary and could be better justified (or referenced). And once this is applied in the AOD retrieval scheme, of course the limit on AOD is clear and you don’t actually need any real data to justify it. The Beer law and the airmass give it straight. The red curve in Fig.12 can be a simple calculation, it doesn’t need data to be supported.
This is maybe not so obvious for the lunar data, especially because the retrieval scheme is not described. Some example of Sun and Moon Vo’s (typical values for a Cimel T) would help to understand the differences, as well as the noise levels. Ultimately the signal-to-noise ratio appears to be more limiting factor in the lunar data than the diffuse light (sky) contribution within the instrument field of view, which is in turn the limiting factor with the Sun (wavelength differences may occur, though).
- Line 267: why not showing 870 or 1020 nm here if those wavelengths are more pertinent to cloud screening?
- Line 274-275: this is not a conclusion of this study. It’s rather a general comment that could be moved to the introduction.
- Line 282: the usable hours of Sun and Moon strongly depend on latitude and season, see comment on section 3.1.1. This part of the conclusions can be enhanced with a global vision (instead of a mid-latitude vision only).

Citation: https://doi.org/10.5194/egusphere-2026-1506-RC2
- AC4: 'Reply on RC2', Joel Schafer, 29 Jul 2026
  
  The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-1506/egusphere-2026-1506-AC4-supplement.pdf
  
  Citation: https://doi.org/10.5194/egusphere-2026-1506-AC4

Peer review completion

AR – Author's response | RR – Referee report | ED – Editor decision | EF – Editorial file upload

AR by Joel Schafer on behalf of the Authors (04 Aug 2026) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (06 Aug 2026) by Daniel Perez-Ramirez

RR by Anonymous Referee #3 (20 Aug 2026)

RR by Anonymous Referee #1 (20 Aug 2026)

ED: Publish as is (15 Sep 2026) by Daniel Perez-Ramirez

AR by Joel Schafer on behalf of the Authors (23 Sep 2026) Author's response Manuscript

Journal article(s) based on this preprint

15 Oct 2026

An assessment of lunar photometry in AERONET

Joel Schafer, Thomas F. Eck, Ilya Slutsker, Pawan Gupta, Mikhail Sorokin, Alexander Sinyuk, and Alexander Smirnov

Atmos. Meas. Tech., 19, 6487–6502, https://doi.org/10.5194/amt-19-6487-2026,https://doi.org/10.5194/amt-19-6487-2026, 2026

Short summary

Joel Schafer, Thomas F. Eck, Ilya Slutsker, Pawan Gupta, Mikhail Sorokin, Alexander Sinyuk, and Alexander Smirnov

Supplement

https://doi.org/10.5194/egusphere-2026-1506-supplement

Joel Schafer, Thomas F. Eck, Ilya Slutsker, Pawan Gupta, Mikhail Sorokin, Alexander Sinyuk, and Alexander Smirnov

Viewed

Total article views: 1,288 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	Supplement	BibTeX	EndNote
803	406	79	1,288	159	67	68

HTML: 803
PDF: 406
XML: 79
Total: 1,288
Supplement: 159
BibTeX: 67
EndNote: 68

Views and downloads (calculated since 26 Mar 2026)

Month	HTML	PDF	XML	Total
Mar 2026	327	153	21	501
Apr 2026	156	73	10	239
May 2026	111	66	15	192
Jun 2026	26	10	4	40
Jul 2026	58	29	6	93
Aug 2026	15	14	3	32
Sep 2026	63	31	14	108
Oct 2026	47	30	6	83

Cumulative views and downloads (calculated since 26 Mar 2026)

Month	HTML	PDF	XML	Total
Mar 2026	327	153	21	501
Apr 2026	156	73	10	239
May 2026	111	66	15	192
Jun 2026	26	10	4	40
Jul 2026	58	29	6	93
Aug 2026	15	14	3	32
Sep 2026	63	31	14	108
Oct 2026	47	30	6	83

Viewed (geographical distribution)

Total article views: 1,262 (including HTML, PDF, and XML) Thereof 1,262 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 26 Oct 2026

Download

The requested preprint has a corresponding peer-reviewed final revised paper. You are encouraged to refer to the final revised version.

Preprint (5608 KB)
Metadata XML

Short summary

A global network of ground-based instruments has been using moonlight to measure atmospheric aerosols at night, producing the world's largest such dataset. By developing a correction method that accounts for the Moon's complex and variable brightness, nighttime measurements now agree with daytime measurements to a high accuracy. This capability is critical because nighttime pollution can differ meaningfully from daytime levels.


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