Review of the manuscript Biologically effective daily radiant exposure for erythema appearance, previtamin D3 synthesis and clearing of psoriatic lesions from erythema biometers at Belsk, Poland, for the period 1976-2023, submitted by Janusz W. Krzyścin, Agnieszka Czerwińska, Bonawentura Rajewska-Więch, Janusz Jarosławski, Piotr S. Sobolewski, and Izabela Pawlak for publishing in Earth System Science Data.

General comment.

The manuscript presents a long (1976−2023) homogenised time series of biologically effective UV daily radiant exposures, obtained from regular monitoring at the Central Geophysical Laboratory of the Institute of Geophysics, Polish Academy of Sciences (IG PAS), Belsk. UV levels included are (i) measured erythemal exposure and (ii) derived cutaneous synthesis of previtamin D3, and clearing of psoriatic lesions.

For measurements, various broadband meters were used along the period treated, thus a series homogenisation has been required. The authors use various methodologies to homogenise series and giving robustness to the dataset.

The authors give links to two datasets corresponding to raw and reevaluated data. Both datasets are easily downloadable and useable, and show the features described by the authors.

In general, the presentation of the manuscript is clear and logical. Overall, the manuscript is well written and informative, deserving for publication.

Specific/technical comments.

Specific and minor comments on formal aspects, mainly asking for explanations and for improving readability:

1. Consider including, in the introduction, and if possible, a short revision for long available series of UV and erythemal data. The reference to Chubarova et al. (2000), which appears in Section 4, can be included there.
2. The different periods and instruments operating along time are repeated several times along the text. Would it be possible to present a table schematising the time periods, instruments, references, the ancillary information used, and methods/models applied?
3. Lines 11-12. I suggest moving the parentheses “(i.e. energy weighted…)” before “reaching the Earth's surface…”
4. Lines 28 and 54-56. Clearness index is usually defined (e.g. Liu and Jordan, 1960) as the ratio between horizontal global irradiance and the extraterrestrial (top of the atmosphere) irradiance. Instead, the authors use the clear sky index: i.e. the ratio between actual horizontal global irradiance and that corresponding to clear sky conditions (which can be simulated). I suggest using the denomination ‘clear sky index’, instead of ‘clearness index’.
5. Line 91. “well-maintained Brewer”. I suggest saying something about the Brewer maintenance and stability.
6. Line 120. What’s the meaning of ‘pre-calibrated’?
7. Line 147. ‘radiance’ Do the authors mean ‘radiation’?
8. Lines 177-178. Add (eryt), (vitD3), (psor) in the Figure caption, as appears in the plot legend.
9. Line 199. Why the authors want to “allow for greater variability in the CC values”? I think the sentences in lines 514-519 do contribute to clarify this question. Thus, I suggest to move that explanation to Section 2.3.2.
10. Line 231. Reference to Outer (2010) appears as ‘den Outer’ in the references list (and in line 58).
11. Lines 247-248. Include in the heading of Table 1 some reference to the Mod1 empirical model.
12. Section 2.4. Why the authors use statistics over relative values (z_i) and not directly over the values? Could this be biasing the results by giving excessive weight to low values?
13. I suggest, when assessing agreement between two series or sets of data, defininig the relative difference by subtracting the reference values in the numerator: this would change the sign of the differences, giving positive values when the tested value overestimates with respect the reference value.
14. Also, I suggest using the term ‘deviation’ instead of ‘error’; this would lead to use MRD, MAD, SD and RMSD instead of MRE, MAE, SE and RMSE.
15. Line 339. The reference to LOWESS has been already given in line 195.
16. Line 609. Reference to Blumthaler et al. (1989) should be before those to Borkowski.

General comments

Most long-term measurements of ultraviolet (UV) radiation at the Earth’s surface started in the late 1980s and early 1990s. The dataset from Belsk discussed by the authors, which started in 1976, is one of the very few exceptions. This dataset is valuable because it includes the period when ozone depletion caused by the raise of chlorofluorocarbons in the atmosphere started to have an impact on UV radiation. To the best of my knowledge, there is no dataset starting before the late 1980s that was obtained with the same instrument and that used consistent quality control procedures. While the Belsk time series is affected by the change of instrumentation, drifts in the calibration, and other instrument artifacts, as described by the authors,  it is one of the few dataset that is suitable for quantifying changes in UV radiation over the period 1976 until today.

To correct for artifacts in their measurements, the authors used comparisons with radiative transfer model calculations and reconstructions from proxy data, e.g., total ozone and cloud attenuation in the visible range. These methods are not ideal for correcting instrument drifts or differences in the characteristics of the instruments used to measure biologically active radiation; however, they are more or less the only two options that are available for obtaining a consistent dataset. I commend the authors for their effort to create a climate data record that is useable for trend detection and other purposes in light of the challenges encountered. I endorse the publication and feel that the journal of Earth System Science Data is an appropriate place to present this important dataset. However, I also feel that the presentation needs to be improved. The manuscript is not an easy read, in particular Section 2.3.4. Readability and usefulness of the article could be greatly enhanced if the authors were to better describe the principles of the correction methods in one or two sentences at the start of each subsection before delving into details. I also find that the nomenclature is sometimes confusing as explained in more detail in my specific comments.

Specific comments

L14, L37: I believe the term “biometer” is only used by Solar Light and Yankee Environmental Systems but no other companies. Since the authors also present data from the original Robertson-Berger meter and the Kipp and Zonen UV radiometer, it would be good to use a more neutral term such as “broadband radiometer with erythemal response”.

L25-28: This should be better explained. First state that two different homogenization methods were applied to the measurements of the various instruments to derive a consistent dataset for the period 1976–2023. Then state that this dataset was further validated by reconstructing independent datasets of DRE and the UV index from proxy data (i.e., total column ozone, aerosol optical depth, and the global irradiance clearness index). When I first read the article it was not immediately clear to me what the “three regression models” entailed.

L68: Please provide a reference for the “Dave-Halpern model”. (It is not clear whether the reference Słomka and Słomka (1985) also includes a description of this model.)

L88 and L188: The UV index is not the midday value of erythemal irradiance. It is erythemal irradiance multiplied with 40 m2/W.

L103: Why “May 1975”? I thought measurements started in 1976 according to the paper’s title?

L138: A symbol for the “daily CI” should be introduced, such as CI_d. I got confused later in the manuscript because I recalled from line 54 that CI is an instantaneous measurement while later in the manuscript all modelling is based on the daily CI.

L147: “To support the quality of the UV observations at Belsk,” > “To validated the corrected UV observations at Belsk”

L166: To improve readability, start this section with describing what the radiative transfer model is used for. For example, state that the TUV radiative transfer model is used to quantify and correct biases in the UV radiation measurements of the radiometers.

L179: Why are you using the monthly mean AOD? Why not higher frequency? Why not use daily AOD on days where AOD is available?

L180-181: “𝑅𝐸_𝐸𝐹𝐹,𝐶𝑙𝑒𝑎𝑟−𝑆𝑘𝑦” > “𝑅𝐸_𝐸𝐹𝐹,CS”. (“CS was used earlier to indicate clear sky, please use consistent acronyms of variables throughout the manuscript).

L193: Regarding “for the days when clear sky conditions can be assumed from the ancillary data.”: Do you mean days that were clear sky from sunrise to sunset? If not, please clarify.

L192-198: The terms “multiplier” and “calibration coefficients” are not clear and this caused confusion when I first read the manuscript. What you define as “raw erythemal data” on line 192 are not really raw data (which would be volts or amps). They are calibrated data (e.g., erythemal exposure or the UV index) affected by drifts and other artifacts. So what you refer to as “multipliers” or “calibration coefficients” are actually correction factors that are applied to calibrated data. I may sound picky pointing this out, but by using more appropriate terms, the readability of the manuscript could be greatly improved. The term “calibration procedure” that you use on line 192 should also be changed to “correction procedure”. To be consistent with this new nomenclature, CC1 and CC2 should also be renamed to CF1 and CF2 for correction factor 1 and 2. (Although this suggestion would be in conflict with using CF for “conversion factor” in Section 2.3.3. So please be creative and find nomenclature that better describes the various factors than that are used in the manuscript.) Also, what does “locally weighted” mean on line 194?

L199: Why would you “allow for greater variability in the CC values”. Ideally, the CC values are the best estimate of the correction for a given day. Why would you like to have “greater variability” for this correction?

L201: “Accordingly, the following conditions were applied for the selection of clear sky sets:” > “The following conditions were applied for the selection of clear sky data used in the two correction methods:”

L207-208: CC2 is based on UVI at noon. If so, why does this depend on sunshine hours for SZA < 85. Whether or not the Sun is shining at times other than noon is irrelevant for the noontime UVI. (If I understand correctly, you only need to ensure that there is clear sky at noon when the comparison between the measurements and model takes place.)

L209-210: I don’t understand this sentence. How can a recalibration in 2011 be informative for a period prior to 1 January 1993? Do you mean that data collected prior to 1993 were assessed in 2011?

L220-225: “current 𝐷 day” > “day D”. (If I understand correctly, D indicates a specific day within the period 1996-2023. So it is not the “current” day.) Also the meaning of the asterisk in D* is a bit murky. You may say that the course of the SZA on a given day of the year is more or less the same in every year. Hence, the conversion factor only considers the TCO for the day in question plus the course of the SZA for the day of the year that corresponds to that day.

L229-230: This sentence should be improved and extended. Please better describe what you did, e.g.: “We developed several regression models from data of the period 2014−2023 by correlating measured radiant UV exposures against UV exposures calculated from proxy data such as TCO and the daily clearness index. We then applied these regression models to proxy data of the entire period (1976-2023) to provide a quality measure of the corrected UV datasets.” Since three regression model are considered, these should also be briefly introduced here so that the reader knows what to expect in the remainder of this section. For example: “The first model (Mod1) is based on clear-sky spectra determined with the RT model discussed in Section 2.3.1 and a cloud modification factor derived from CI data. The second and third model (Mod2 and Mod2) are based on TCO and short-wave irradiance (G) data evaluated on a monthly basis. TCO and short-wave irradiance were either taken from observations at Belsk (Mod2) or ERA5 reanalysis (Mod3).”

L233: The CI was defined on line 55 as the "quotient of the all-sky global solar irradiance (GSI) at  the surface and the corresponding synthetic clear-sky value to account for combined  cloud/aerosol scattering effects." Hence, CI is a function of the instantaneous irradiance not a daily value, as CI(D) implies. So what is CI(D)? is it a daily average? Clear definitions of CI, CI(D) and CMF should be provided. (If CI(D) is the daily average, CI(D) is not a good acronym as such a notation would mean “instantaneous clearness index as a function of day D” with D being the argument of CI. CI_D would be a better variable name. (Underscore indicates a subscript.))

L301: Figure 3a does not show a ratio, as the caption states, but is a correlation or scatter plot.

L330: Again the use of the term “calibration coefficients” is confusing in this context. Please use “correction factors” or something similar. (Thank you for using the term “correction method” on line 337 and not “calibration method”!)

L336: “are proposed (Sect. 2.3.2) using” > “were proposed in Sect. 2.3.2 using”

L341: Please specify range of the "former" period.

L355-356: Describe better and use different colors as those used in panel (a). What's shown are not differences between calibration coefficients but the difference “(Modelled RE - Observed RE) minus (Modelled UVI - Observed UVI)”.

L401: Regarding: “to fill gaps in the proxy data”: Do you mean fill gaps in the measurements of the biometers? (I thought the ERA5 reanalysis data _are_ the proxy data.)

L518: I would not conclude that the CC1 and CC2 methods "very different." In fact, they are quite similar. Just one is based on daily exposure while the other one on the noontime UVI. If the responsivity of a biometer is off by a given factor, it would affect daily and noontime values equally. It is therefore not surprising that CC1 and CC2 results are similar.

L548: Did the instrument used by Chubarova also have an erythemal response or did it have a constant response over the 300-380 nm range? If the latter, the signal would be dominated by wavelengths in the UV-A; hence, the sensitivity to variations in ozone would be minimal.

L560-568: I would think that Sections 6 should come before Section 5.

Technical comments

I presume the article will be prove-read by a copy editor. I therefore keep my comments regarding language to a minimum.

Always add "radiation" after "UV" or define “UVR” as "ultraviolet radiation" (e.g., in line 52, change “surface UV modelling” to “modelling of surface UV radiation”) This comment applies to many instances.

L2, L166, 177: Change “erythema appearance” to just “erythema”. The term “erythema appearance” is not commonly used.

L3: “from erythema biometers” > “derived from broadband radiometers with erythemal response” (The term erythema biometers does not describe the instrument and is not widely used.)

L17: “the appearance of erythema,” > “erythema”

L40: penetrate > reach

L41: “were destroyed” > “would be destroyed”

L73. Please explain acronym “IG PAS” (presumably Institute of Geophysics of the  Polish Academy of Sciences, but this is not mentioned).

L76: Mod > Model

L107: “MED=210 J_eryt m-2,” > “1 MED=210 J_eryt m-2,”

L126: proved > proven

L137: “CI is a commonly used” > “The CI is a commonly used”

L167: Move “irradiance at noon in day D” before “Ir_EFF”

L172: “𝐼𝑟(𝜆, 𝑡)” should be “𝐼𝑟_CS(𝜆, 𝑡)”. Change “in time 𝑡 for the wavelength” to “at time 𝑡 and at wavelength”

L192: “The calibration procedure” > “The correction procedure

L204: “difference between observed sunshine duration and theoretical one” > “difference between the observed sunshine duration and the theoretical one”

L205: “0.5 hour as for higher” > “30 minutes. This limit was chosen because broadband UV measurements at larger SZAs …”

L232: “that the erythemal” > “the erythemal”

L253: “in the UV explaining variables X” > “in variables X that affect UV radiation”

L281: Opening parenthesis missing in the formula for F.

L283: “(for 𝑅𝑘+1 <0, F=1):” > “If Rk+1 is smaller than 1, F is set to 1.”

L304: (CC ver. 1) and What does “(CC ver. 1)  mean? This was not defined earlier. Is this equal to CC1? Please use consistent nomenclature!

L307: What do you mean with “CF values” here? CF values were earlier defined as conversion factors from erythema to other weighting functions. I suspect you mean CC1 values here. Again, please use consistent nomenclature!

L311: “as shown by the linear regressions close to the 1-1 perfect agreement line in the three scatter plots (Fig. 5).” > as shown in the scatter plots of Fig. 5, which indicate that Brewer and biometer data cluster about the ideal 1:1 line.

L322: “used in routine” > “measurements used in routine”

L369: see > Note

L427: “determination coefficients” > “coefficients of determination”

L436: thesis > hypothesis

L497 comes > are

L521: “as shown by the comparisons with BS64 data” > “because they agreed well with the BS64 data”

L598-599: the title “References” appears twice.