General comments

The authors analyze a very long (the world’s longest as they claim) homogenized series of measured erythemal solar irradiance. Then, for the same period (1976 - 2023) they reconstruct the series using different proxies to determine the contribution of different factors to the observed trends. Although the manuscript is well structured and well written, there are some major issues that must be considered by the authors prior to the submission of a revised version of the paper.

• To homogenize the series the authors have scaled the measurements using the reconstructed series of the noon erythemal doses. In my opinion, the scaling should be performed using low-turbidity days to avoid biases due to the impact of changes in the prevailing aerosol species. Gradual changes in aerosol composition would affect optical properties such as the single scattering albedo (SSA) and the Angstrom Exponent (AE), that can affect UV doses and their changes have not been taken into account. Furthermore, the characteristics of different sensors possibly affect the results, even after the scaling to the modelled doses. For example, imperfect angular response, or temperature dependence are possible different for different sensors. If such effects have not been considered they increase the uncertainty in the homogenized series. There should be at least some discussion regarding the remaining uncertainties after the homogenization.
• The proxies that have been used for the series reconstruction do not always correspond to UV-B (i.e., the part of the solar spectrum that mostly contributes to the erythemal doses). For example, the relationship between the AOD at 340 nm and the AOD below 315 nm depends on the AE (practically from the aerosol species) which has been assumed to be equal to 0. The SSA has not been considered and can also have significant impact on the UV trends. The effects of clouds depend on their type and properties. There should be at least some discussion on the uncertainties related to these factors.

Specific comments are provided below.

Specific comments

L8: “measured erythemal “ instead of “erythemal”. There are reconstructed series that are longer.

L10: “observations period” instead of “observations”

Introduction: In additions to being the main source of vitamin D, exposure to UVR has other positive effects. E.g.,:

https://www.tandfonline.com/doi/full/10.4161/derm.20013

https://www.mdpi.com/1660-4601/13/10/1028

Section 2.2: How was the clearness index “translated” to clearness index for ERE? Please provide more details. Furthermore, what about other aerosol properties (e.g., SSA, AE) that affect ERE? Furthermore, please provide a reference for the erythema actions spectrum that has been used to calculate ERE.

L184-186: Given that the AE can practically range from ~ 0 (e.g., for dust) to ~2 (e.g., for biomass burning aerosols), and that the greatest contribution to ERE comes from wavelengths at 306 – 308 nm, there can be a difference of up to ~20% between AOD at 340 nm and the AOD at wavelengths that contribute more significantly to ERE. There should be at least some discussion about that. Furthermore, during the cold period, can changes in surface albedo have played a role? Does the assumption of a default surface albedo introduce any uncertainty?

L233: “monitored” instead of “monitoring”

L277-278: What does 3%p means?

L387: Delete “much”

The paper ‘Trends in the erythemal radiant exposure from re-evaluated measurements (1976−2023) with biometers at Belsk, Poland, and their sources from corresponding ozone, aerosol, and cloud observations ; by Czerwinska et al., investigates a time series of erythemal solar irradiance 1976-2023 in Belsk, Poland and analyses trends and potential factors influencing it. This topic is very relevant for a better understanding of the evolution of solar uv radiation received at the surface, in particular if there is an increasing or decreasing trend and which factors possibly influence this evolution. The strength of the paper is that the influencing factors like total ozone, aerosol load, cloud coverage are taken into account by several statistical means and time scales. The manuscript is fitting into the scope of Atmospheric Chemistry and Physics. The manuscript is well written and I can recommend publication for ACP after some revision, described below.

General comments:

The authors claim that their time series of homogenized erythemal uv dose is the longest globally. This should be substantiated by giving at least some references of other relevant long time series and respective references.

The authors should give for their instruments used for deriving the Erythemal Radiant Exposure (ERE) the wavelength range measured, mentioning also ‘broadband’ (or spectral in case), this could e.g. be integrated in table 1. It should be indicated if these sensors used by the authors have been calibrated regularly against reference standards.

The authors explain (also via the co-article in ESSD) that the quality of the ERE time series of the KZ sensors was accessed by checking with ERE observations of the Brewer #64. The Brewer has however another uv range measured. How was the ERE therefore derived in order to assure comparability? Further, for the Brewer#64: in section 2.2, Ancillary data, it is not included in the TCO3 database? Why?

The introduction should also include a WMO definition of the erythemal (uv) irradiance and its relation to the commonly communicated uv index.

Specific comments:

Line 50: the Brewer #64 spectrophotometer is mentioned here, but not a bit further above in the list of instruments used for monitoring important parameters; The Brewer should also be listed there briefly (since when, where?). Or was its data not used any further?

Lines 76-77: ‘the months 1985 (June and July) were exluded’ …:  what does this mean? Only for the year 1985, or for all years these two months? If June and July have been excluded for all years, then it would be necessary to do the same calculations once with including them in order to check the uncertainty budget. If only 1985, then why this year?

Section 2.2 Ancillary Data:  It should be more detailed how the time series for these individual influencing factors have been assured to be homogeneous; e.g. the TCO3 between Dobson and satellite observations, and the G_o values (if ECMWF or from MERRA-2)? How was it decided to take the one or the other? In case of AOD from Aeronet – which data level has been taken? SunDur: ‘percentage of daily duration’: daily duration means from sunrise to sunset, or fixed day duration?

Lines 101-105: The formulation why to take years 1999/2000 as turning point can be improved. Later on, the authors give more details (lines 135-139), but at this place here, it is a bit confusing (especially the formuation ‘somewhere in between’). When looking at the values in table 3, the R2 value for year 2000 is always lower (in one case equal) than the R2max for another year – although in line 137 it is stated to choose the year with R2 max. It feels thus that the choice for years 1999/2000 needs an improved motivation. E.g. volcano eruptions (Pinatubo, …), enter into force of the Montreal Protocol, instrumental issues, further statistical change point analysis….

Lines 184-188: In addition to the comments of the other reviewer: How valid is the assumption that AOD at 340 nm is representative for the UV-B range? There should be some discussion on it, including some references to measurements of AOD in the UV-B.

Line 360, table 9: Would it be possible to increase readability? It is not intuitive to read connected numbers so easily

Line 379: ‘neighbouring stations: do the authors mean Hradec Kralove and Lindenberg? Then this should be clarified (‘these neighbouring …’)

Lines 383-385: For the mentioned numbers for Reading and Sodankylä – please clarify to which reference these belong

Discussion section on uv trends: Are there references for trend analyses reaching further than the ones mentioned?  E.g.: https://doi.org/10.5194/acp-22-12827-2022; https://doi.org/10.5194/acp-21-18689-2021; https://doi.org/10.1007/s43630-024-00658-8; https://doi.org/10.1002/joc.7803

Lines 402-411: this paragraph fits rather to the conclusions, as it describes the results rather than to discuss them

Lines 418-419: Is there a source reference for the statement that the atmosphere over Belsk was getting cleaner?

Line 423: ‘red circles and blue crosses’: these are not the physical entities behind; please write out the meant individual factors/proxies.

Lines 435-437: why should there no further increase of UVR be expectable? Total ozone is highly variable, also atmospheric aerosol load and how these factors will change is not determined, in particular taking climate change into account. Climate change is also expected to impact on cloud cover. Overall warmer temperatures and a purer atmosphere support less cloud cover.

Technical comments:

Line 88 and in figure 1 caption: define CI for the clearness index

Line 126:  please change to ‘The largest differences …’

Line 282: ‘… in the per cent …’ please skip the ‘the’

Line 288ff: ‘ozone & aerosols’; better write ‘combined ozone and aerosol impact’ or similar

Line 324: in brackets: (approximately between 0.8 to 0.9)

Line 325: skip ‘per cent’ (% already there)

Line 378: please change to ‘Republic and Lindenberg’

Line 383: please rephrase: ‘mainly related to the changes in aerosol load, albedo and cloud cover …’