the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
UV Index monitoring and forecasting in Cyprus: the effect of ozone, aerosols and clouds
Abstract. Surface ultraviolet (UV) radiation is strongly influenced by clouds, aerosols, and total column ozone, making accurate operational forecasting challenging in regions with highly variable atmospheric conditions. This study evaluates UV Index (UVI) variability and forecasting over Cyprus using more than one year of observations from a five-station ground-based monitoring network. Spectral measurements from a Bentham spectroradiometer were combined with broadband UV observations, all-sky imagery, AERONET aerosol measurements, and satellite ozone products to evaluate the regional Cyprus Erythemal Radiation Forecasting System (CERYFOS) and the Copernicus Atmosphere Monitoring Service (CAMS) corresponding product. Ground-based observations showed a strong seasonal cycle across Cyprus, together with regional differences associated with altitude and local atmospheric variability, particularly at the mountainous Kyperounda station. Both forecasting systems reproduced observed UVI well under typical atmospheric conditions. For the all sky conditions, CERYFOS and CAMS achieved RMSE values of 0.72 and 0.60 UVI, respectively, decreasing to 0.47 and 0.40 UVI under clear-sky cases. In contrast, performance deteriorated under cloudy conditions, with RMSE increasing to 1.65 and 1.12 UVI, respectively, identifying cloud representation as a major source of forecast uncertainty. Ozone-related uncertainties generally produced relatively small UVI deviations, whereas aerosol errors became increasingly important under enhanced aerosol loading, as particularly evident during the 23–24 July 2025 wildfire, underestimated aerosol loading resulted in forecast errors exceeding 4 UVI units. These results demonstrate the importance of accurate cloud and aerosol representation and coordinated ground-based observations for improving operational UVI forecasting in the Eastern Mediterranean.
Competing interests: At least one of the (co-)authors is a guest member of the editorial board of Atmospheric Measurement Techniques for the special issue “Sun-photometric measurements of aerosols: harmonization, comparisons, synergies, effects, and applications”.
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.- Preprint
(2554 KB) - Metadata XML
-
Supplement
(702 KB) - BibTeX
- EndNote
Status: final response (author comments only)
- RC1: 'Comment on egusphere-2026-6018', Anonymous Referee #1, 11 Nov 2026
-
RC2: 'Comment on egusphere-2026-6018', Anonymous Referee #2, 12 Nov 2026
The paper ‘UV Index monitoring and forecasting in Cyprus: the effect of ozone, aerosols and clouds’ by Charalampoud et al. investigates observed surface UV radiation and derived UV index for five stations in Cyprus and compares it to two forecasting systems, the Cyprus CERYFOS and the Copernicus CAMS system. This topic is relevant for improving operational UVI forecasting in the Eastern Mediterranean. The combination with influences of total ozone, clouds and atmospheric aerosol on UV radiation adds scientific value to the manuscript. The methodology is solid and the authors investigate and assess uncertainties related to SZA, clouds, ozone and aerosol conditions and their impact on the forecasted UVI. The manuscript is well fitting into the scope of the AMT-ACP special issue (Sun-photometric measurements of aerosols: harmonization, comparisons, synergies, effects, and applications). The manuscript is well written and I can recommend publication after some revision, described below.
General comments:
The direct comparison of forecasted UVI between all-sky, clear-sky and cloudy sky with measurements is only made for the Limassol site. I understand this is because there is also an All sky imager located and thus the data set can be well distinguished in subsets. Nevertheless, I think the paper can gain in relevance if also for the other 4 sites in Cyprus with UVI measurements such subsets were made. Is there a possibility to characterize the cloud or cloud-free state for these sites? E.g., by collocated global radiation data? Or by using satellite imagery?
Specific comments:
Lines 19, 20: ‘local atmospheric variability’, ‘typical atmospheric conditions’ – what is meant with these expression? A brief description would help.
Line 24: ‘aerosol errors’ – please be more precise on the aerosol parameters concerned
Line 46: ‘under high aerosol loading’ – what are high aerosol loadings, please be precise
Lines 59-60: which UVI levels exactly? Numbers?
Line 74: reference Gröbner et al., 2005 not in references
Lines 84ff: the authors introduce the Cyprus Solar Network, but their forecasting system CERYFOS is then just mentioned, without – briefly – introducing when it was started and e.g., some references. Please give some brief introductory details on CERYFOS here.
Line 93: ‘solar geometry’ – please describe what is meant? Solar Zenith Angle?
Table 1 : please give also the starting date of respective observations
Line 164: ‘expanded uncertainty +/- 5%’: I suppose this is for W/m2 ?
Section 2.2.2: Broadband UV radiometers: As I understand, a full uncertainty assessment is underway. I’d nevertheless like to ask if there is any reference yet? In addition, in line 211, the authors mention that the 1min values were aggregated to hourly values for comparisons with the Bentham (15miin). As UVI can be highly variable, have the authors also investigated to aggregate to the 15min intervals of the Bentham?
Lines 205-206: ‘intercomparisons’ – does this mean all the five broadband UVE radiometers or only the one at the Bentham station?
Section 2.3: Satellite-based TOC: please give information how the TOC values per station have been retrieved with respect to the satellite grid resolution.
Line 290: weblink to the forecasting site not accessible (12/11/2026)
lines 309-313: please describe the acroynyms GOME-2, TEMIS, DEM
lines 320-321: please refer for the CMF to your formula 5
lines 343-345: it seems like only a part of a phrase – please re-phrase
lines 345-348 are similar to lines 358-360 – please aggregate
lines 372-373: how could the authors investigate the spatial variability with the one site Limassol?
Line 391-392: sentence ‘ To characterize … ‘ is double, its content was already mentioned bit before.
Line 398, equation 4: are here also the local noon UVI values used? Please clarify.
Lines 501-508: The authors mention ‘several periods characterized by reduced UVI under cloud-free conditions’ and describe then the July 2025 wild fire event and mention the atmospheric enhanced aerosol loading. What about the variation in TOC ?
Further, it is interesting that the highest UVI are measured around and after the summer solstice. Do the authors have some ideas on this ?
Line 540, Figure 3: It would be good to see how the five station UVI measurements compare at low UV radiation levels. Would it be possible to add a respective figure for the whole year in the supplements?
Lines 592-593: The Aeronet NIC site is collocated with the ATHA UV site – then why comparing it here with the AMX site?
Lines 608-613: this statement was already similarly mentioned bit above – please consider some aggregation.
Line 618: Figure 4 shows the CMF for 6 stations – the Agia Marina Aeronet station included; why then not also the Nicosia Aeronet site?
Line 710: larger variability at larger SZA – could it also be a straylight effect? It should be mentioned and/or discussed
Line 791, table 4 caption: ‘the reported metrics mean bias, median bias, standard deviation (Std), RMSE’ are only in the text, not in the table.
Line 823, section 3.3.3: why is only AOD at 380 nm used and not (also) AOD at 340 nm which is more relevant for the UV?
Line 840, Figure 10 and section 3.3.3: please mention again, to clarify, that the Aeronet data used here are for the Limassol station
Line 860, table 5 caption: ‘the reported metrics mean bias, median bias, standard deviation (Std), RMSE’ are not given
Lines 881-882: the SSA and asymmetry parameters were extrapolated to 340 nm; Can the authors give some explanation, references that also for these parameters, such an extrapolation is valid. Because for AOD, this is commonly done, but how is this for SSA and asymmetry parameter?
Lines 897-898: CAMS ‘overestimates SSA in JJA’, but CAMS absorbs too strongly during dust events – is this not contradictory, as the main dust event period is during JJA?
Lines 922-927: It would be interesting, to have an idea on the frequency of such events with high aerosol loading, leading to such high AOD discrepancies, transferring then to high UVI forecast discrepancies
Line 956: ‘aerosol condition forecast’ : please clarify here what exactly are these aerosol conditions
Technical comments:
Line 60: ‘JJA’ – please describe
Line 299: ‘AFGL’ please describe
Line 343: (eq. 2)
Line 675, Figure 6: I suggest to increase the font size of the legend boxes
Line 707, reference to figures S2, S3: could within these figures the font size be enlarged?
line 878: ‘asy’: although the abbreviation has been explained earlier, here it is better to write its whole name, at least here in the beginning
Citation: https://doi.org/10.5194/egusphere-2026-6018-RC2
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 331 | 140 | 40 | 511 | 63 | 32 | 44 |
- HTML: 331
- PDF: 140
- XML: 40
- Total: 511
- Supplement: 63
- BibTeX: 32
- EndNote: 44
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
I enjoyed seeing the data from Cyprus, especially the wildfire event, and learning more about the ground-based measurement network on the island.
Unfortunately the link to the UVI Nexus Hub web application, where the CERYFOS UVI forecasts can be found, was not accessible at the time of my review.
The manuscript has clear potential, and I would like to suggest a few points that should be addressed before publication.
Major points
Bentham UVI time series (Fig. 2): October 2024 to September 2025.
Comparison of ground-based UVI measurements (Fig. 3): high-insolation period 2025 (exact date range not stated).
Cloud modification factor (Fig. 4): period unclear (October 2024 to September 2025?).
Modelled vs. measured UVI (Fig. 6): is the data hourly or local noon (as stated in the caption)? Which period?
SZA-binned UVI differences (Figs. S2 and S3): hourly data? Which period? The sample size N is very similar for all SZA bins in each panel, which does not seem consistent with hourly data. What about data from higher SZA bins (morning and evening data)?
Modelled vs. measured UVI (Fig. 6): The sample size N differs between CERYFOS and CAMS, so the statistics cannot be compared directly. A comparison on a common set of data pairs would solve this. This is especially relevant for cloudy situations (N = 531 for CERYFOS and N = 923 for CAMS).
The ozone differences are calculated with respect to OMI ozone, which is provided once a day. For a period of one year I would therefore expect about N = 365 values, but N = 6720 is stated in the legend. Could the authors clarify this?
The abstract suggests a comparison of all five ground stations with the forecasts. Besides the data from Limassol, it would be very valuable to also use the broadband data from the other four stations in the comparison with the forecasting models CERYFOS and CAMS. This would also allow the authors to support the statement in lines 574-578 that forecasting is more demanding at Kyperounda with analysis results. This would add great value to the manuscript.
The impact of the clear-sky filtering method (selection of clear-sky data points based on the obstruction of the solar disk alone) is not discussed in chapter 3.2.1. It would be valuable to analyse it, e.g. with a stricter filter (no clouds at all).
The conclusions chapter could be strengthened, as it currently mainly repeats statistics given earlier.
Minor points
Statistics in tables could be moved to the supplement, leaving only the most important statistical values in the main text.
Statements such as the +-5 min window are given several times in the first half of the manuscript.
Chapters 2.2 to 2.5 could be shortened, as other publications describe the instrumentation and data and the manuscript is already quite long.
Some references cited in the text are missing in the reference section: Gröbner et al. (2005), Skamarock et al. (2019), Bhartia (2007), Fountoulakis et al. (2024), Kazadzis et al. (2016), Levelt et al. (2018), Papetta et al. (2026) and Bais et al. (2015).
Some references do not exactly match the entry in the reference list: Zerefos et al. (2023), Antón et al. (2015), NOAA (1981)
As already noted in the text, the values in column two (Mean/Median Day-to-Day Change [UVI]) are close to zero by construction, so they add little information and I would suggest removing them from the manuscript.
The maximum values (columns 5-7) are also of limited significance, as they may largely reflect changing cloudiness. Detector problems or data gaps may also have contributed to these maximum values.
Statistics for clear-sky data points could be added to remove the impact of cloudiness.
Please state the number of local-noon values per site.
The maximum increase of the Bentham and the broadband device at Limassol differ significantly (5.96 and 3.99), although these co-located measurement devices are described as behaving similarly.
In chapter 3.3.4, SSA and asy values at 340 nm are used in the analysis. It would therefore be more consistent to also use AOD340 in chapter 3.3.3.
Technical remarks