Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula

Čížková, Klára; Láska, Kamil; Metelka, Ladislav; Staněk, Martin

doi:https://doi.org/10.5194/egusphere-2022-688

Preprints

https://doi.org/10.5194/egusphere-2022-688

Preprints

07 Sep 2022

| 07 Sep 2022

Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula

Klára Čížková, Kamil Láska, Ladislav Metelka, and Martin Staněk

Abstract. This study aims to assess response of spectral UV radiation to different atmospheric and terrestrial factors, including solar zenith angle, ozone, cloud cover and albedo, in southern polar environment. For this purpose, 23260 individual spectra (300–363 nm), obtained by the B199 Mk-III Brewer spectrophotometer at Marambio Base, Antarctic Peninsula Region, over the period 2010–2020, were studied. A neural network model was developed to investigate the effects of the explanatory variables at individual wavelengths. SZA proved to be the most important parameter, followed by cloud cover, TOC, and surface albedo. The relative SZA effect is greatest at the shortest wavelengths, where a 1° decline in SZA results in a 6–18 % increase in UV irradiance (305 nm). Also TOC affects particularly the short wavelengths, while at 305 nm, a 10 DU decrease in TOC causes a 7–13 % increase in UV irradiance. In some seasons (e.g., 2011–2012, 2014–2015, 2018–2019), the large-scale ozone holes caused the UV irradiance at very short wavelengths peak in spring, whereas in other seasons (e.g., 2010–2011, 2012–2013), the maxima at all wavelengths were recorded in summer. The effect of cloud cover was strongest near the fully cloudy sky, and in summer months. Albedo affected the UV irradiance only weakly, its effect was strongest in early spring and late fall.

Received: 23 Jul 2022 – Discussion started: 07 Sep 2022

Download & links

Preprint (PDF, 7913 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 (7913 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

18 Apr 2023

Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula

Klára Čížková, Kamil Láska, Ladislav Metelka, and Martin Staněk

Atmos. Chem. Phys., 23, 4617–4636, https://doi.org/10.5194/acp-23-4617-2023,https://doi.org/10.5194/acp-23-4617-2023, 2023

Short summary

Klára Čížková et al.

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2022-688', Anonymous Referee #1, 13 Oct 2022

Anonymous Referee

Comments to the manuscript of Klára ÄíÅ¾ková et al., Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula

The study aims to assess the relative effect of factors affecting surface spectral radiation, including SZA, total ozone, cloudiness, and albedo. The dataset is from Brewer spectrophotometer measurements at Marambio Base from the Antarctic Peninsula. Spectral measurements at those latitudes are rare and thus the spectral UV time series 2010-2020 is very valuable, and the manuscript gives a good overview of the statistical distribution of the data. The idea of studying the effect of each UV affecting parameter is not new as such, and that has been studied in several publications previously. However, in this study, the use of neural network makes the study interesting and unique, in addition to the Antarctic location. Unfortunately there is one clear mistake in the applied inputs to the model: the surface albedo data can not be taken from the OMUVB product. In the OMUVB product, the albedo is not the actual measured albedo, but it is the albedo climatology which is used to calculate the OMI UV data. I think the paper can not be published before this has been taken into account, either by excluding the part of the study related to albedo effect or by using other data reflecting the actual albedo situation of the site. I have also a concern on the quality and homogeneity of the time series, as there was so few calibrations (see specific comments here below). In addition I don’t understand what is the purpose of Figure 11 and text related to that figure (please see specific comments here below).

I am not familiar with ANN modeling, but I have the impression, that some additional information could be given related to the ten different models which were used (Line 140). How did these models differed from each others.

Specific comments:

Line 28, Reference UNEP2010 could be changed to a more recent one, eg., UNEP2019

EEAP. 2019. Environmental Effects and Interactions of Stratospheric Ozone Depletion, UV Radiation, and Climate Change. 2018 Assessment Report. Nairobi: Environmental Effects Assessment Panel, United Nations Environment Programme (UNEP) 390 pp. https://ozone.unep.org/science/assessment/eeap

line 30: → incident short wavelength UV irradiance. Or change the wording to describe that it prevents the short wavelength UVC and UVB part of the UV spectrum.

Line 33: ….

Please specify in which part of the globe? Or global mean?

Line 88: Please specify the method for calibration against the reference instrument. Usually a Brewer is calibrated using 1kW or 200W lamps. Please specify also the traceability of the irradiance scale. Any comparison with the PMOD-World reference QASUME spectroradiometer? Either between the IOS and QASUME or between Marambio’s Brewer and QASUME? Only two calibration for 11 years of measurements is very little. There is a possibility that the response of the instrument has changed unexpectedly between the calibrations. Do you have any records to check if the instrument has been stable over the years? How did you take into account the change in calibration: linear interpolation between the two calibrations or stepwise? How much was the difference between the two calibrations? Did you perform any final calibration in 2020? Please explain these points in the text.

Line 96: Which method did you use for detection of spikes and wavelength shifts? Please add references.

Line 120: What did you use as inputs to LibRadtran? Please add the info in the text. And how did you calculated the CMF?

Line 123: This is not true. The albedo is the albedo climatoly used in calculation of the OMUVB product. Please see my General comments.

Line 154: Do you mean the 80% of data was within +-25%?

Line 156: What is R2?

Lines 156-157: What do you mean by shares variability and shared variability? Do you mean shared variance, covariance? Please use other wording.

Line 177: I suggest that you include the info of how many order of magnitude the UV irradiance changes between e.g., 300 nm and 400 nm.

Line 179: ……..it is steeper in low-SZA months…….. Explain why (longer atmospheric path – more ozone absorption)

Line 185: What do you mean with “overall median”. Yearly median? Something else?

Line 192: The vertical profile of atmospheric ozone affects the absorption in the optical path. This info could be added somewhere, and possibly discussed.

Lines 193-197: I think you should, in a couple of sentences, describe the effect of the Brewer-Dobson circulation and the lack of sunlight in winter → ozone accumulates in polar region, but there is no sunlight for the photochemical ozone destruction which then leads to high TOC values at the end of the winter. You should also explain the year to year variation in Antarctic ozone depletion, otherwise it is quite strange that you find within during the same week the highest and the lowest ever measured TOC (6 November 2011, 3 November 2015).

Line 199: Where is the site of Lachlan-Cope (2010) located?

Lines 200-2004: I am not convinced why do you compare with these sites, and about the reasons for the differences. Is there differences in land-sea surroundings, glaciers ...topography, other meteorological reasons, typical synoptic scale phenomena?

Lines 228-229:

How about tropospheric O3, SO2 ? What do you mean by “

Lines 245->end of the section. Please specify what do you mean by “very high UV irradiances”. Very high compared to what? Or higher than a certain limit value?

Line 277: This term doesn’t exist.

Line 285: Should this be in high-SZA..?

Lines 309-317 I think this discussion should be excluded if actual albedo values are not used (See general Comments)

From line 318 until the end of the section. I don’t understand what has been done. Why didn’t you keep in the model the same values than in the observation (Figure 11)?

Figures 7-9 caption: Include the information if the effect is the median effect, and if yes, to which quantity do you make the change (actual value during the measured spectrum?).

Figure 10. I think you should exclude this plot, as the albedo is not the actual one. You can not make any conclusion of it’s influence.

Citation: https://doi.org/10.5194/egusphere-2022-688-RC1
- AC1: 'Reply on RC1', Klara Cizkova, 29 Nov 2022
  
  Thank you for the review and the comments! Please find the detailed answers in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2022-688-AC1
RC2:
'Comment on egusphere-2022-688', Anonymous Referee #2, 19 Oct 2022

Review of the study entitled “Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula” by Klára ÄíÅ¾ková et al.

I read an interesting article which aims to investigate the effect of solar zenith angle, ozone, cloud cover and surface albedo on spectral UV radiation at Marambio Base, Antarctic Peninsula. UV irradiance measurements come from a double Brewer spectrophotometer for the period 2010-2020. The effects of the different parameters on surface UV irradiance are studied using a neural network model that has been developed for this purpose. My recommendation for this article is to accept for publication after clarifying better what Figure 11 aims to show, as I had also recommended during my quick review, and after clarifying the neural networking explained in section 2.4.

Specific comments:

Lines 139-161: I find difficult to understand how the ten models were built and how the best neural network model was selected each time. Given that most studies use multiple regression modelling to quantify the contribution of each atmospheric parameter, lines 140-142 trigger the question how much different would the calculations from a multiple regression model be? A supplement with explanations on the neural model procedures, and comparison with estimations from a multiple regression model would help.

Low albedo case (fig. 11d): the Obs. Alb. value is indeed low (0.37), but the Mod. Alb. value is 0.81, which is not low. Please check.

Citation: https://doi.org/10.5194/egusphere-2022-688-RC2
- AC2: 'Reply on RC2', Klara Cizkova, 29 Nov 2022
  
  Thank you for the review and the comments! Please find the detailed answers in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2022-688-AC2

Interactive discussion

Status: closed

RC1:
'Comment on egusphere-2022-688', Anonymous Referee #1, 13 Oct 2022

Anonymous Referee

Comments to the manuscript of Klára ÄíÅ¾ková et al., Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula

The study aims to assess the relative effect of factors affecting surface spectral radiation, including SZA, total ozone, cloudiness, and albedo. The dataset is from Brewer spectrophotometer measurements at Marambio Base from the Antarctic Peninsula. Spectral measurements at those latitudes are rare and thus the spectral UV time series 2010-2020 is very valuable, and the manuscript gives a good overview of the statistical distribution of the data. The idea of studying the effect of each UV affecting parameter is not new as such, and that has been studied in several publications previously. However, in this study, the use of neural network makes the study interesting and unique, in addition to the Antarctic location. Unfortunately there is one clear mistake in the applied inputs to the model: the surface albedo data can not be taken from the OMUVB product. In the OMUVB product, the albedo is not the actual measured albedo, but it is the albedo climatology which is used to calculate the OMI UV data. I think the paper can not be published before this has been taken into account, either by excluding the part of the study related to albedo effect or by using other data reflecting the actual albedo situation of the site. I have also a concern on the quality and homogeneity of the time series, as there was so few calibrations (see specific comments here below). In addition I don’t understand what is the purpose of Figure 11 and text related to that figure (please see specific comments here below).

I am not familiar with ANN modeling, but I have the impression, that some additional information could be given related to the ten different models which were used (Line 140). How did these models differed from each others.

Specific comments:

Line 28, Reference UNEP2010 could be changed to a more recent one, eg., UNEP2019

EEAP. 2019. Environmental Effects and Interactions of Stratospheric Ozone Depletion, UV Radiation, and Climate Change. 2018 Assessment Report. Nairobi: Environmental Effects Assessment Panel, United Nations Environment Programme (UNEP) 390 pp. https://ozone.unep.org/science/assessment/eeap

line 30: → incident short wavelength UV irradiance. Or change the wording to describe that it prevents the short wavelength UVC and UVB part of the UV spectrum.

Line 33: ….

Please specify in which part of the globe? Or global mean?

Line 88: Please specify the method for calibration against the reference instrument. Usually a Brewer is calibrated using 1kW or 200W lamps. Please specify also the traceability of the irradiance scale. Any comparison with the PMOD-World reference QASUME spectroradiometer? Either between the IOS and QASUME or between Marambio’s Brewer and QASUME? Only two calibration for 11 years of measurements is very little. There is a possibility that the response of the instrument has changed unexpectedly between the calibrations. Do you have any records to check if the instrument has been stable over the years? How did you take into account the change in calibration: linear interpolation between the two calibrations or stepwise? How much was the difference between the two calibrations? Did you perform any final calibration in 2020? Please explain these points in the text.

Line 96: Which method did you use for detection of spikes and wavelength shifts? Please add references.

Line 120: What did you use as inputs to LibRadtran? Please add the info in the text. And how did you calculated the CMF?

Line 123: This is not true. The albedo is the albedo climatoly used in calculation of the OMUVB product. Please see my General comments.

Line 154: Do you mean the 80% of data was within +-25%?

Line 156: What is R2?

Lines 156-157: What do you mean by shares variability and shared variability? Do you mean shared variance, covariance? Please use other wording.

Line 177: I suggest that you include the info of how many order of magnitude the UV irradiance changes between e.g., 300 nm and 400 nm.

Line 179: ……..it is steeper in low-SZA months…….. Explain why (longer atmospheric path – more ozone absorption)

Line 185: What do you mean with “overall median”. Yearly median? Something else?

Line 192: The vertical profile of atmospheric ozone affects the absorption in the optical path. This info could be added somewhere, and possibly discussed.

Lines 193-197: I think you should, in a couple of sentences, describe the effect of the Brewer-Dobson circulation and the lack of sunlight in winter → ozone accumulates in polar region, but there is no sunlight for the photochemical ozone destruction which then leads to high TOC values at the end of the winter. You should also explain the year to year variation in Antarctic ozone depletion, otherwise it is quite strange that you find within during the same week the highest and the lowest ever measured TOC (6 November 2011, 3 November 2015).

Line 199: Where is the site of Lachlan-Cope (2010) located?

Lines 200-2004: I am not convinced why do you compare with these sites, and about the reasons for the differences. Is there differences in land-sea surroundings, glaciers ...topography, other meteorological reasons, typical synoptic scale phenomena?

Lines 228-229:

How about tropospheric O3, SO2 ? What do you mean by “

Lines 245->end of the section. Please specify what do you mean by “very high UV irradiances”. Very high compared to what? Or higher than a certain limit value?

Line 277: This term doesn’t exist.

Line 285: Should this be in high-SZA..?

Lines 309-317 I think this discussion should be excluded if actual albedo values are not used (See general Comments)

From line 318 until the end of the section. I don’t understand what has been done. Why didn’t you keep in the model the same values than in the observation (Figure 11)?

Figures 7-9 caption: Include the information if the effect is the median effect, and if yes, to which quantity do you make the change (actual value during the measured spectrum?).

Figure 10. I think you should exclude this plot, as the albedo is not the actual one. You can not make any conclusion of it’s influence.

Citation: https://doi.org/10.5194/egusphere-2022-688-RC1
- AC1: 'Reply on RC1', Klara Cizkova, 29 Nov 2022
  
  Thank you for the review and the comments! Please find the detailed answers in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2022-688-AC1
RC2:
'Comment on egusphere-2022-688', Anonymous Referee #2, 19 Oct 2022

Review of the study entitled “Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula” by Klára ÄíÅ¾ková et al.

I read an interesting article which aims to investigate the effect of solar zenith angle, ozone, cloud cover and surface albedo on spectral UV radiation at Marambio Base, Antarctic Peninsula. UV irradiance measurements come from a double Brewer spectrophotometer for the period 2010-2020. The effects of the different parameters on surface UV irradiance are studied using a neural network model that has been developed for this purpose. My recommendation for this article is to accept for publication after clarifying better what Figure 11 aims to show, as I had also recommended during my quick review, and after clarifying the neural networking explained in section 2.4.

Specific comments:

Lines 139-161: I find difficult to understand how the ten models were built and how the best neural network model was selected each time. Given that most studies use multiple regression modelling to quantify the contribution of each atmospheric parameter, lines 140-142 trigger the question how much different would the calculations from a multiple regression model be? A supplement with explanations on the neural model procedures, and comparison with estimations from a multiple regression model would help.

Low albedo case (fig. 11d): the Obs. Alb. value is indeed low (0.37), but the Mod. Alb. value is 0.81, which is not low. Please check.

Citation: https://doi.org/10.5194/egusphere-2022-688-RC2
- AC2: 'Reply on RC2', Klara Cizkova, 29 Nov 2022
  
  Thank you for the review and the comments! Please find the detailed answers in the attached document.
  
  Citation: https://doi.org/10.5194/egusphere-2022-688-AC2

Peer review completion

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

AR by Klara Cizkova on behalf of the Authors (19 Dec 2022) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (23 Dec 2022) by Stelios Kazadzis

RR by Anonymous Referee #3 (19 Jan 2023)

Suggestions for revision or reasons for rejection

This is a review of a revised version of a manuscript titled “Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula” by Čížková et al. I have also reviewed the response of the authors to the comments of two reviewers pertaining to the original version of the manuscript. I confirm that that the authors have addressed the reviewers’ concerns appropriately. Unfortunately, I have found additional major flaws in the revised manuscript, which must be addressed before the manuscript can be accepted for publication. Most importantly, data of the Brewer spectrophotometer and the complementing model calculations should have been reported as “spectral irradiance”, not “irradiance”. Furthermore, data presented in the manuscript are too low by at least two orders of magnitude. Most data therefore have to be reprocessed. In addition, data presented in Figures 7–9 show large artefacts related to the artificial neural network model. Most notably, Figure 8c indicates that UV irradiance depends on total column ozone at wavelengths larger than 350 nm when in fact (and confirmed by my own calculations) the ozone absorption cross section in this wavelength range is too small to have a noticeable effect on UV irradiance. The dataset is worth publishing, considering that there are only a few sites in Antarctica that provide spectral UV data. However, my “major” and “minor” comments should be addressed first, and the language should be improved also. I have included several pages related to language with suggestions to improve the text and make it more readable.

***Major comments
**Data are not provided in the correct quantity and the magnitude of the results is incorrect
All results are provided as “irradiance” in units of mW m-2. A Brewer MK III measures spectral irradiance, not irradiance, in units of mW m-2 nm-1 (with a spectral resolution (bandwidth) of 0.5 nm according to the authors; although, to my knowledge, the resolution of this instrument is 0.6 nm). If the authors provide results in irradiance instead of spectral irradiance, the wavelength interval over which their spectral data were integrated to get irradiance has to be specified. Furthermore, irradiances presented by the authors seem to be off by several orders of magnitude. For example, for a solar zenith angle (SZA) of 60°, I would expect a spectral irradiance of 335 mW m-2 nm-1 at 363 nm for a spectroradiometer with a 1 nm wide triangular slit function. The “irradiance” data presented by the authors in Figure 4 max out at about 0.8 mW m-2. This is different from the correct result (presuming that irradiance data were integrated over a wavelength interval of 1 nm) by a factor of more than 400. Even if data had only been integrated over 0.5 nm (the resolution and sample-step of the Brewer) instead of 1 nm, as in my calculations, the results would be off by two orders of magnitude. As a result of these errors, all subsequent results presented in absolute terms (irradiance in mW m-2) and model results (which are tied to measurements) in Figure 7–10 are incorrect. All absolute values in the manuscript have to be recalculated to correct for these errors.

**The artificial neural network (ANN) is not described.
It is stated in line 156 that the ANN model is a “perceptron” model, but no other details are provided. Please provide information on this model, either a reference or a description on how it works. Without any detail, it is just a black box. Journal papers like this require a description that would allow the reader to reproduce these model calculations.

**The spectral UV measurements are insufficiently described.
For example, how many spectra were measured per day? Are these spectra distributed equally throughout the day (otherwise the average and median of daily measurements would be skewed). What is the uncertainty of the measurements? Have the measurements been independently validated, for example, by participating in an intercomparision campaign with a reference spectroradiometer.

**Figures 7-9 show artifacts of the ANN model, so some features presented in these figures are not real, and some artefacts lead to wrong conclusions.
Results should be double-checked with a radiative transfer model (which the authors apparently have because they calculate CMFs). It would be simple to model the UV irradiance as a function of SZA for each month for a fixed TOC and CMF. It would be much more convincing to have results based on a physically correct model than the author's ANN model. The following features are likely artefacts of the ANN model:
In Figure 7a, UV irradiances do not asymptotically approach zero for large SZAs but seem to have an offset. In Figure 7b, there is a spurious maximum at about SZA=45 degree. In Figure 7d, large, wavelength-dependent fluctuations are apparent at small wavelengths in particular for March and April, etc.
The greatest artefacts are apparent in Figure 8c. Because of the steep decline of the ozone absorption cross section in the Huggins band, the effect of TOC on UV irradiance becomes very small for wavelengths larger than 340 nm, and in particular for wavelengths larger than 350 nm. Figure 8c contradicts the diminishing effect of changes in TOC on UV irradiance with increasing wavelength. Here, a 10 DU change in TOC has the largest effect on UV irradiance at 363 nm (in January). Furthermore, Figure 8d shows unrealistic spikes in the relative change at about 303 and 305 nm for March and April. The relative change in Figure 8d seems to be constant (albeit small) for wavelengths larger than ~337 nm. Instead the relative change should continuously decrease between 337 and 363 nm.

To check my assertions, I modeled UV spectra for SZA=60 degree and TOC of either 300 or 310 DU (i.e., an increase by 10 DU) using the ozone absorption cross-section by Molina, which is provided up to 350 nm. As expected, the relative change decreased with increasing wavelength and was below 0.02% for wavelengths larger than 345 nm. The absolute change peaked at about 317 nm at a irradiance of 2.1 mW m-2 nm-1, decreased steadily with increasing wavelengths from there onward, and was below 0.02 mW m-2 nm-1 above 348 nm. The difference that I calculate with my radiative transfer model are greatly different both in magnitude and shape of the difference shown in Panel (c), confirming my conclusion that the results of the ANN model are incorrect.

In summary, Figure 8c gives a false impression of the effect of changes in TOC on UV irradiance.

Figure 9c shows unrealistic fluctuations below 310 nm.

In general, I find the “absolute” panels (c) in Figures 7–9 not very helpful because the structure is dominated by the Fraunhofer lines. The relative changes shown in panels (d) are really the interesting ones and the authors should focus their discussion on these panels.

**Important literature is not cited and results are ignored
The assessment in line 61 that “a complex evaluation of long time series of solar UV irradiance spectra from Antarctica is still missing.” is incorrect. For example, measurement of spectral UV irradiance at the South Pole, McMurdo / Arrival Heights, and Palmer Station, have been continuously performed since the early 1990s as part of the National Science Foundation’s and NOAA Antarctica UV Monitoring Network, e.g., https://gml.noaa.gov/grad/antuv/. Numerous publications are based on these data and these publications also include quantitative assessments of the effects of ozone, albedo, cloud cover, and other factors on the measured UV spectra. The authors should consult the list of references in Bernhard and Stierle, 2020 (which they have cited), in particular #21, 23, 24, 28, 30, and 43, as well as https://doi.org/10.1029/2004JD004937, which includes a very detailed analysis on the wavelength dependence of cloud attenuation, and the variability in UV radiation caused by variations in total ozone at the South Pole.

***Minor comments
The introduction contains much general information that could be cited and does not have to be included. So the introduction could be shortened substantially
L21: Include reference to support “since its discovery in 1801.”
L34: Jovanović et al., 2019 is a rather obscure publication. Please add a more standard one.
L37: the paper by Bernhard and Stierle does not report on positive trends in UV irradiance for the month of September. (Positive trends were reported for other months, but not for September).

41–42: The wavelength ranges for UV-C, UV-B and UV-A radiation should be:
UV-C: 100–280
UV-B: 280–315
UV-A: 315–400

L63: I am not convinced that there is a gap. The way SZA and ozone affects UV radiation are well understood. The effect from clouds is more difficult, but assessing the complex effects of clouds with cloud cover data from satellites covering a larger area will do little to capture these intricate effects. Please see also my last “major” comment.

L67: Regarding: “between the southern polar vortex and UV irradiance reaching the Antarctic continent”: The paper describes measurements from one particular site, which is on an island and not the Antarctic continent. So these measurements can hardly be representative for the vast Antarctic continent.

L93: It was already stated in the previous sentence that the instrument was calibrated in 2012 and 2016. So what does "the spectrum was calibrated in 2012, 2016” add here?

L96: What does R6 and R5 mean? What was ratioed? What are Dead time and Run stop tests? Why is “Dead” and “Run” capitalized?

L100: What wavelength range does "very short wavelengths" refer to? (The range is explained further down. This explanation should be moved up).

L106: Where wavelengths shifts just analyzed or were they actually corrected?

L117: According to Figure 4, bottom, cloud cover at Marambio is the norm. So what TOC data were used on cloudy days?

L127: Is “cloudiness” equal to cloud cover? How is cloud cover expressed in this reanalysis data set? Is it in oktas or percent? Is it the fraction of clouds within the 0.25 x 0.25 degree pixel? Since Marambio is on a small island, are the ERA5 data representative? The optical thickness of clouds is at least as important as cloud cover. Hence, it seems that cloud cover from ERA5 reanalysis data is not a good quantity to assess the effects of clouds on UV radiation.

L128: Regarding “the best correlation (r = 0.26) with”. Best relative to what? Is it indeed r or should it be r^2? If it were r, r^2 would be 0.068, which means that only 6.8% of the variability in the CMF can be explained with ERA5 cloud cover data. This is a rather small fraction, confirming that the ERA5 reanalysis data are not very useful to describe cloud cover at Marambio, which also impacts the significance of results presented later. This should be discussed.

L130: “then the weighted mean using the clear-sky intensities as weights was derived for each spectrum.” Is difficult to understand. Please rephrase and add detail.

L138: How representative is the OMUVB albedo climatology for the relative small footprint of the observation site considering that there is both ocean and land in the OMUVB pixel? Were there any albedo measurements at Marambio Base that could be used to validate the OMUVB albedo climatology?

L153: You only considered the ANN and a regression model. Why didn't you use a physical correct radiative transfer model to detangle the effect of the different explanatory variables considering that all important input variables for such a model were available?

L158: What is the difference between testing and validation?

L173: Delete: “while it was greater at shorter wavelengths, likely due to the smaller range of the data.” This does not make sense to me. My guess is that R^2 is greater at small wavelengths because those depend more on ozone, which (anti)correlates much better with UV radiation than cloud cover.

L183: What 9 months?

L197: Delete: “the UVB region”. 330 nm is not in the UV-B range.

L211: I don’t understand “relative differences gradient”

L212: Regarding: “In October, when the UV irradiance is lower than the overall median, the difference decreases with decreasing wavelengths, while in November, when the median irradiance already exceeds the overall median, the difference increases more steeply.” Looking at Figure 4c, the UV irradiance is higher (not lower) than then overall median in October and the difference increases (not decreases) with decreasing wavelength.

L218: Karhu et al. (2003) or Koo et al. (2018) are not good references to cite here. I suggest to cite either the 2022 WMO ozone assessment (https://csl.noaa.gov/assessments/ozone/2022/) instead, which will be published soon, or the 1998 version.

L225–229: While this is more or less correct, the explanation misses the point why both ozone minima and maxima can be observed in November. The main reason is that Marambio is sufficiently far away from the South Pole so that the station can be either below the ozone hole or outside its perimeter where ozone is typically high in November.

L269: Attenuation of UV-B irradiance by thin (e.g., cirrus) clouds can be much less than 30%, in particular if snow is on the ground.

L296–300. I think you mean Fig. 7–9, not Fig. 7-10, in the paragraph.

L299: The sentence “of the relationships of the explanatory variables and UV irradiance, given that all other variables are fixed to their monthly medians.” is confusing. The sentence implies that there are other variables, in addition to explanatory variables and UV irradiance. Better: “Fig 7 shows the relationship between SZA and UV irradiance with the other explanatory variable fixed to their monthly mean. Similar relationships between TOC and UV irradiance and between cloud cover and UV irradiance are shown in Figs. 8 and 9, respectively.”

The captions of Figure 7–9 should be rearranged. I suggest (for Figure 7): “Modelled relationships between UV irradiance and SZA for different months. Panel (a) and (b) show the modelled UV irradiance at 305 nm (a) and 340 nm (b) as a function of SZA. Panels (c) and (d) show the change in spectral UV irradiance resulting from an increase in SZA by 1° in absolute (c) and relative (d) terms. Similar captions should be chosen for Figures 8 and 9.

L328: The vertical profile becomes only important for SZA larger than about 75 degrees, when the Umkehr effect starts to become apparent. So the ozone profile does not plan a “substantial” role, unless the SZA is very large.

L340: Another important reasons why clouds have less influence in Antarctica compared to mid-latitude sites is that high surface albedo prevailing over the Antarctic continent greatly reduces cloud effects due to multiple scattering between clouds and the snow-covered surface. However, this may be of less importance at Marambio compared to the Antarctic continent.

Figure 10: I find this figure rather confusing and not very helpful. It is obvious that measurement and model don't agree when the model is run with input parameters differing from the actual situation. I think the main point that the authors like to make is that low ozone values can be compensated by large SZAs (e.g., the red lines in Panel b). I would remove the figure from manuscript but let the authors decide.

L405: I find it rather strange that the results of nine of the ten ANN models are basically identical but one (ANN01) is rather different. I would have expected a continuous distributions. It would be good if the authors could better explain what’s different about ANN01. It seems to me that the training dataset of model ANN01 included outliers.

***Comments to language
**Comments to language, general
The word “while” is used improperly throughout the document. “While” contrasts two different things, like “whereas”. For example, one could say: irradiance at 305 nm is small while it is large at 340 nm. “While” is often used in the text instead of “for example”, which is an incorrect meaning.
Change “in average” to “on average” throughout.
Don’t use the expressions “high SZA” and “low SZA”. These expressions are confusing because a reader associates “high” with the Sun being high in the sky while the opposite is the case. Use “large SZA” and “small SZA” instead.
Change “Out of” to just “Of”
Always place a space between value and unit. (It is done most of the time, but not always.)
Change “Huggins belt” to “Huggins band”
**Comments to language, specific
L7: “assess response of spectral UV radiation to different atmospheric” > “to assess the response …” or better: “assess the dependence of spectral UV radiation on different…”
L8: “in southern polar” > “in the southern polar”; delete “individual unique” as this is obvious
L11: “the resolution of 0.5 nm.” > a resolution of 0.5 nm.” Also please double-check that this is correct for your instrument. To my knowledge, the spectral resolution of a Brewer MK III is 0.6 nm.
L12: Define “TOC”
L13: decline > decrease
L13: Regarding the sentence: “Also TOC affects particularly the short wavelengths, while at 305 nm, a 10 DU decrease in TOC causes a 7–13 % increase in UV irradiance.”
Specify what you consider “short” wavelengths. The word “while” implies that 305 nm is a long wavelength because “while” is a word contrasting two different things. So you could say: TOC affects wavelengths below about 340 nm. For example, at 305 nm a 10 DU decrease in TOC causes a 7–13 % increase in UV irradiance.”

L17: Specify what “very high” means (e.g., the upper 10% of the distribution)

L22: Delete “the thorough”
L23: accounted > attributed
L25: "lead to melanoma" should refer to humans, not "other organisms”
L26: “catalyses vitamin D creation” > “leads to the production of vit D”
L29: Cite https://doi.org/10.1038/s41586-021-03737-3 in support of “terrestrial plant productivity”
L31: “short-wavelength UV irradiance” > “from reaching the surface”
L32: “in the 1980s” > “in 1985”
L33: “many events have taken place to eliminate the ozone depletion,” > “ many efforts have been made to reduce ozone depletion”
L33: “the 1987 Montreal Protocol and its numerous amendments” > “through the passing of the Montreal Protocol in 1987 and subsequent amendments to this landmark treaty.”

L35: “in September” > “for the month of September”

L37: “ozone hole still” > “the ozone hole still”

L38: Delete “the rare”

L43: Remove “as it is illustrated even by the possible division at 315 nm” The division is indeed somewhat arbitrary, but the publications cited do not “illustrate” this.

L48: Indicate that NH means nitrogen hydrogen

L49: “but due to their respective abundance” > “but due to the difference in their respective abundances”

L53: precise > large

L54: “the solar UV spectra” > “solar UV spectra”

L55: delete “already”

L57: Delete: ”especially without further processing”

L58: Delete "the UV Index” (The UV Index is not an action spectrum)

L66: “artificial neural network modelling (ANN)” > “artificial neural network (ANN) modelling”

L69: Avoid strings of nouns: “solar UV spectral irradiance observation” > “observations of spectral solar UV irradiance”

L79: doesn’t > does not

L80: “all year long” > “at any time of the year”

L88: “has been performed” > “was performed”

L89: travel > travelling

L92: “of HG and CD > “of Hg and Cd” or better “mercury and cadmium”

L102: “weak transmitted information affected” > “spectral UV irradiance at the Earth surface, which is affected”

L107: “The 23 260 of the spectra,” > “The subset of 23260 spectra that passed quality control were used for this study and paired with explanatory variables” (if that’s what you mean)

L109: “It can be seen there were several data gaps, of which the longest occurred “ > “It can be seen that there were several data gaps, the longest of which occurred”

L113: “has been paired with following explanatory” > “was paired with the following explanatory”

L115: “belong to” > “are”

L119: “Therefore, more solar spectra could be matched” > “Therefore, several solar spectra were sometimes matched”

L120: “provided it was taken within the 60 minute time distance and it was the closest observation.” > “provided that they were taken within 60 minutes of the closest ozone observation”

L134: “(further).” > “(see Section XX).”

L153: “further in this section).” > “further below”

L162: “more information is included in Appendix” > “see Appendix”

L166: “However, even in” > “However, even for”; “carried on: median bias values” > carried out: the median bias between measurement and model values”

L167: delete “then they”

L168: “tackled” > “removed”

L169: “fitted between” > “agreed to within”

L170: “from approximately 310 nm,” > “for wavelengths longer than approximately 310 nm,”; “modelled data was within approximately” > “measured and modelled data agreed within approximately”

L172: “determination coefficient” > “coefficient of determination”

L174: “shared variance” is an uncommon term. Use “coefficient of determination” instead.

L177: “Out of the four explanatory variables, always only a single one was left to its original value, while the three other variables were fixed to their monthly medians” > “Of the four explanatory variables, one was selected and retained at its original value while the three other variables were fixed to their monthly medians. The procedure was then repeated for each of the four variables.”

L195: “increased” > “increases”

L200 “of the Sun and the studied site” > “of the Sun at Marambio Base”

L202: “increase in median irradiance varies” > “increase in median irradiance per nm varies”

L205: Delete “precise”

L214: “The wave-like Huggins belt” > “The wave-like structure in the Huggins band”

L217: “for Antarctic” > “for the Antarctic”

L222: “conditioned” > “depend both on”

L270: Delete “likely”

L277: “which may lead to its decline compared to snow cover or glaciated areas” > “which will lead to lower UV radiation compared to snow cover or glaciated areas”

L307: “indirectly proportional” > “inversely proportional”

L385: “wavelengths, when at” > “wavelengths. At”

L390: “at very short wavelengths is most visible.” > “is most visible at very short wavelengths.”

Hide

ED: Reconsider after major revisions (19 Jan 2023) by Stelios Kazadzis

AR by Klara Cizkova on behalf of the Authors (24 Feb 2023) Author's response Author's tracked changes Manuscript

ED: Referee Nomination & Report Request started (02 Mar 2023) by Stelios Kazadzis

RR by Anonymous Referee #3 (16 Mar 2023)

RR by Anonymous Referee #4 (17 Mar 2023)

ED: Publish as is (18 Mar 2023) by Stelios Kazadzis

AR by Klara Cizkova on behalf of the Authors (22 Mar 2023) Author's response Manuscript

Journal article(s) based on this preprint

18 Apr 2023

Assessment of spectral UV radiation at Marambio Base, Antarctic Peninsula

Klára Čížková, Kamil Láska, Ladislav Metelka, and Martin Staněk

Atmos. Chem. Phys., 23, 4617–4636, https://doi.org/10.5194/acp-23-4617-2023,https://doi.org/10.5194/acp-23-4617-2023, 2023

Short summary

Klára Čížková et al.

Viewed

Total article views: 386 (including HTML, PDF, and XML)

HTML	PDF	XML	Total	BibTeX	EndNote
267	103	16	386	4	5

HTML: 267
PDF: 103
XML: 16
Total: 386
BibTeX: 4
EndNote: 5

Views and downloads (calculated since 07 Sep 2022)

Month	HTML	PDF	XML	Total
Sep 2022	87	34	5	126
Oct 2022	54	17	5	76
Nov 2022	31	9	4	44
Dec 2022	23	7	0	30
Jan 2023	17	6	0	23
Feb 2023	26	18	2	46
Mar 2023	20	6	0	26
Apr 2023	9	6	0	15

Cumulative views and downloads (calculated since 07 Sep 2022)

Month	HTML	PDF	XML	Total
Sep 2022	87	34	5	126
Oct 2022	54	17	5	76
Nov 2022	31	9	4	44
Dec 2022	23	7	0	30
Jan 2023	17	6	0	23
Feb 2023	26	18	2	46
Mar 2023	20	6	0	26
Apr 2023	9	6	0	15

Viewed (geographical distribution)

Total article views: 386 (including HTML, PDF, and XML) Thereof 386 with geography defined and 0 with unknown origin.

Country	#	Views	%

Latest update: 18 Apr 2023

Download

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

Preprint (7913 KB)
Metadata XML

Short summary

The study deals with ultraviolet radiation in southern polar conditions, where ozone depletion occurs each spring. For the first time in the region, a ten years long time series UV spectra from Marambio Base, Antarctic Peninsula, has been studied, with a focus on the changes of UV radiation at different wavelengths and the effects of atmospheric and terrestrial variables. At the very short wavelengths, the effect of ozone and its deficiency was clearly observed.


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