In this study by Chesnoiu et al. the variability of surface solar irradiance and its components (direct and diffuse) is investigated for the period 2010-2022 over Lille, France, relying on ground-based measurements and radiative transfer simulations. Based on the classification of the sky conditions (for clouds and aerosols) they quantified the contribution of the different parameters on the variability and trends of the different solar irradiance components, and they obtained also climatologies for their site. The objectives of the study are quite straightforward and are addressed by a thorough analysis. The surface solar radiation climatology and trends that the authors provide here for Lille considering also the atmospheric parameters that can impact the calculated changes is of significance for assessing and understand changes in surface solar radiation. I consider the topic and results of this manuscript to fit the scope of ACP. However, I have some general and major comments (please see below 1-5) which should be addressed prior to publication.

1) My major comment concerning this study has to do with Section 4. Authors should rename the section, correct it and state clearer the objectives regarding this analysis. The analysis performed is a quantification of the direct (scattering and absorption) impact of aerosols in downwelling surface solar irradiance. Changes in downwelling surface solar irradiance due to aerosols calculated using eq. 29 are always negative (attenuation) due to their direct interactions of incoming solar radiation which is of relevance for surface related applications like solar energy as stated in the manuscript. The relative change (expressed in %) in downwelling surface solar irradiance due to aerosols presence was calculated with respect to an aerosol-free atmosphere using eq. 27. However, the radiative effect due to aerosol-radiation interactions REari according to IPCC report formerly known as direct radiative effect (DRE), is the change in radiative fluxes caused by combined scattering and absorption of radiation by anthropogenic and natural aerosols (Boucher et al., 2013). DREs are climate related quantities which are calculated at surface and top of the atmosphere (TOA) using net fluxes (downwelling minus upwelling) for shortwave and longwave radiation, for clear sky and all skies conditions, in order to assess the warming of cooling of the earth-atmosphere system.  Authors should address which is the objective of this study, make the appropriate changes in section 4 including related references where applicable.

Boucher, O., D. Randall, P. Artaxo, C. Bretherton, G. Feingold, P. Forster, V.-M. Kerminen, Y. Kondo, H. Liao, U. Lohmann, P. Rasch, S.K. Satheesh, S. Sherwood, B. Stevens and X.Y. Zhang, 2013: Clouds and Aerosols. In: Climate Change 2013: The Physical Science Basis. Contribution of Working Group I to the Fifth Assessment Report of the Intergovernmental Panel on Climate Change [Stocker, T.F., D. Qin, G.-K. Plattner, M. Tignor, S.K. Allen, J. Boschung, A. Nauels, Y. Xia, V. Bex and P.M. Midgley (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA.

2) It is stated at Lines 273-274 that AOD440, AOD550, AE440-870 and relative humidity are the “remaining inputs” to SOLARTDECO. I think that this gives the wrong impression that these are the inputs in RT, while these values are used to select and adjust the new optical properties for the specific aerosol mixture. Please clarify and adjust 2.3.1 and 2.3.2 accordingly. For example, in Lines 343-344 it should be clarified that it is not only the fact that Mie calculations are time consuming that were not performed for every RT simulation, but the fact that inversion data are not always available, which has already stated earlier in the manuscript. In addition, while it is described in detail how the new aerosol optical properties are derived for each simulation, it is not clear the vertical structure of the aerosol layer (only notes in Lines 654-655 and 672-673) and how the extinction is scaled to measured aerosol load in the total column. Please clarify which are the parameters to be set (are those in Table 4?) apart from defining the new aerosol optical properties based on the Mie calculations.  Another confusing part is Lines 383-384. SSA, g and ff are “estimates of optical properties derived from SOLARTDECO”? How this information is consistent with what is stated in Sections 2.3.1 and 2.3.2 regarding the inputs in RT?

3) Consider to move 3.2.1 and 3.2.2 to the methodology Section.

Specific comments

Line 10: Change “all sky” to “all skies” throughout the manuscript.

Line 47: I am not following those "increase" and "decrease" descriptions inside the parenthesis where are referred to?  

Line 49: I suggest including also here the importance of direct normal irradiance to concentrating solar power systems providing also references (e.g. Sengupta et al., 2021, https://www.nrel.gov/docs/fy21osti/77635.pdf)

Lines 271: Please remove the 500 in “of respectively 407 and 209 500 ppmv”.

Lines 273-274: These sentences are confusing, regarding the geometry. The only geometry to be determined is the sun position through solar zenith angle and regarding viewing angle  the whole dome is considered, right? It is stated explicitly in Line 283 that the “horizontal irradiances” were calculated, so please clarify if the other geometries are important at this part since no computations for tilted surfaces were performed.

Line 281: Please consider the change from “of the incoming and outgoing spectral irradiances” to “of the incoming and outgoing spectral solar irradiances”

Line 367: These are absolute differences? Please, clarify.

Line 439: Is 435 W/m2 the correct number for clear-sky? The grey line is below 400 in Fig. 4 (h).

Line 451: Is 69% correct in summer for all-sky conditions according to Fig.4 (e)?  

Lines 653-654 and 658-662: It would be helpful to provide also those mean values used as refence.

Line 660: Remove “logically”.

Figure 7:  Last 3 lines of caption need to be clarified better.

Line 724: F_CSUN here is BHI_CSWC?

Technical corrections

Line 386: Replace 2020 with 2010.

Figure 4: In (e), (f) and (h) percentages that reflect the contribution of the DHI to the overall mean yearly GHI are missing. For (b) and clear-sky this is “blue line” or green? In addition in the 6^th line AERONET is twice.

Lines 507-508: Change the color of the lines insides parenthesis according to Fig. 5b and 6b.

Line 623: Replace “dFclear/dt” with dFi/dt

Line 626: In eq. 22 probably this “Fclear” is F?

Line 791: Word “surface” is twice

Figures 7, 8, 9: Should be enlarge and brown columns better solid than shaded.

Review of paper "Influence of cloudy/clear-sky partitions, aerosols and geometry on the recent variability of surface solar irradiance’s components in

northern France" by G. Chesnoiu et al.

The paper is dedicated at assessing the role of aerosol and clouds in modulating diffuse and global solar irradiance components at Lille.

Identifying different cloudiness conditions, separating and quantifying effects have been and are challenging objectives.

The combination of selection methods which use measurements of direct normal and diffuse horizontal irradiances are applied to identify different cloud and aerosol conditions.

Based on the applied selection methods, the authors assess the role of clouds and aerosol on variability and modulation of global and direct horizontal 

irradiances.  Radiation transfer calculations, and simplified schemes for aerosol characterization based on AERONET observations are used to assess the role played by different parameters in the solar radiation modulation.  Inter-annual variability, especially in spring and summer, is also investigated together with main driving factors.

The paper is very long and incorporates many different methodological aspects, data, and results.   The results are interesting, and the

separation of the different conditions and factors is worth the attempt.  However, the description of some of the applied procedure would need a clearer

presentation and some assumptions need to be better justified and discussed. The impact of the implied uncertainties should also be 

assessed and discussed.

As said, the paper is very long and addresses different topics.  The authors might consider focussing on a smaller number of topics, and/or

separating the content into two different papers. 

Specific details are given below.

l. 101: what is intended with "relatively stable"?  Please, add information  on the calibration changes, and if any calibration correction or other adjustement was applied.  This is important information when a long term data record is being examined,  such as it is the case of this paper.

Section 2.2.1.  In my opinion, the presentation of the applies algorithms should be improved.  Maybe, the addition of a flux diagram might help.

In particular, the method used to seperate clear sky and clear sun conditions should be better justified and discussed.  This is at the basis of most of the

following analyses and requires a stronger verification.  In particular, the method by Batlles et al. (2000) is correctly classified by Guyemard et al. (2019)

among those for "cloudless sky" (clear sky detection methods of the first kind), which are based on solar irradiance measurements.  Thus, the use of the difference between the cases selected by Batlles et al. (2000) as "clear sun" (l. 216) requires in my opinion a specific discussion and justification.  The cases falling in category CSWC also strongly depend on the accuracy of this algorithm.   The need of a stronger justification is also supported by the relatively poor performance found in section 2.2.2 (e.g., 67% precision for the Batlles et al.  method).  Maybe, the adoption of a "clear sun" selection method based on DNI

measurements (see, e.g., Guyemard et al., 2019) might be explored. 

Section 2.2.2.  Also with respect to the previous comment, results of the performance analysis need more discussion.

The performance analysis was carried out considering only two months of 2018.  Was there a reason why January and May were chosen? 

Secondly, the selection obtained for the Batlles et al. (2000) method (clear sun) does not appear to produce very convincing results: there is about 13% of wrong determinations.  How this is taken into account in the following analyses?  How much it may impact the obtained results on cloud forcing in the different situations?

As a test case, I would suggest making calculations for January and May, 2018, based on classifications obtained from the visual analysis of sky imager data,

and comparing results with those obtained by the automatic algorithms.  This might provide an indication on the possible impact of the clear sun selection method.

Section 2.3.2. The presentation of the used aerosol models is not very clear, and a more detailed description is needed.  Various aspects appear confuse to me, and need additional clarification. The definition of some variables is lacking.

I am listing below some questions that in my opinion need clarification; however, some of these may be due to the limited understanding of all the steps used in the method I could obtain after reading the section several times.  Also in this case, maybe a flux diagram may help understanding the applied procedures. The author should state clearly when AERONET measured values and/or simulated data are used in the different steps, and what parameters are used in the simulations.

It is not clear to me how the different size distributions are chosen starting from the AERONET inversions of cases falling in the 60 different classes.  

Do the authors take an average distribution over those retrieved from AERONET observations in the corresponding class?  

I also do not understand the use of ff and cf in equations 6 and 7.  If ff is the fine mode fraction applied to the AOD (eq. 6), in my opinion the same 

factor can not be applied in the same way to AE (eq. 7).  By combining a given fine and coarse mode fraction of AODfine and AOD coarse produces an AEtot which is different from ff AEfine + cf AEcoarse.  Also, please specify how Cext and Csca are derived (I assume they are calculated for the

specific fine and coarse size distribution and refractive index identified for each class) and calculated; is the aerosol number concentration

included in the formula?  How is it derived?  NCfine and NCcoarse are derived from AERONET inversions (l. 339).  If I understood well, they were derived using the size distribution retrieved by AERONET, which may be different from the average ditribution assumed for the specific class in the Mie calculation for the LUT. In this case, the use of AERONET values for NCfine and NCcoarse in the calculated optical properties may produce incorrect results.

l. 383-385: I could not understand what aerosol optical properties were in the end used in the simulations.  Is the method of section 2.3.2, or a different set of values (see also l. 275-278: please, explain how this is made)?

Minor comments are below:

l. 46-47: The sentence "depending on their optical properties, aerosols and clouds influence incident radiation by altering both the direct (increase) and diffuse (decrease) components" is unclear (in particular "increase" and "decrease").

l. 47-49: solar concentration systems rely mainly on the direct component.  This may be mentioned as an additional supporting motivation.

Sections 2.1.2-2.1.5: are the various measurements synchronized?  What is the uncertainty on the time determination of the various instruments?  Synchronicity

is crucial when dealing with clouds, and the use of data/images acquired at different rates and possibly different times may impact the results.

l. 107: according to the manufacturer, the spectral range defined by the 50% points is 200-3600 nm for CMP22, and 200 to 4000 nm for CHP1.

l. 109: irradiance measurements are available at 1-minute tome resolution.  Are these individual measurements or an average over a defined time interval?  If an average, is the standard deviation also acquired?

   

l. 125: "important air mass" is not clear.  What is intended? Relatively large AOD?  In the literature there is a definite threshold on AOD at 440 nm needed to obtain reliable retrievals of SSA

l. 149-150: how wind speed and direction affect simulations of surface solar irradiance?

l. 225-226: how time time delays between irradiance measurement and image is taken into account? 

l. 321: using relative humidity as one of the factors for aerosol classifications implies that the aerosol properties are essentially determined by the mixed layer properties, which is plausible at a continental site like Lille, where however sea breeze might still have some influence, producing a vertical differentiation in the aerosol characteristics.    Are there evident dependencies of aerosol optical properties on surface relative humidity?   Are there previous studies supporting this classification scheme?

l. 246-249: it is not clear to me the advantage of using +/-30 minutes with respect to sunrise and sunset with respect to selecting an appropriate solar zenith angle. This makes probably more difficult a comparison between data obtained with different daily averaging methods.

Figure 4: the legend defines "AERONET reference" what is called "overall" AOD in the caption.  I suggest using the same notation.

l. 422-423: it is not possible to determine the ciclonic/anticiclonic condition from the wind flow direction alone. This should be better explained.

l. 430-432: this is not convincing.  Also airmasses from other directions appear to be possibly influenced by marine aerosol.

   

l. 437: "minimum" istead of "minimal"

l. 478: please, define the imits of the spring and summer periods used in the analysis (different seasonal separations are found in the literature)

l. 486-488: does the statistical significance test take into account the uncertainty in the determination of the different conditions?  I think these trends should be seen with some caution due to the uncertainties associated with the determinations of the specific conditions (in particular occurrence of clear sky with clouds conditions).  Also in the conclusions, possible effects of the uncertainties on trend determinations should be mentioned (l. 1026-1029).

l. 590 and 599: all the terms right of sigma should be included in parentheses

l. 618-619; see previous comment

l. 654: Do the author use a logarithmic aerosol vertical profile, or a single layer at a specific height?  Please, specify what is intended with aerosol layer height.

   

l. 808: using the same symbol for the absolute and the relative direct radiative effect is misleading.

   

Section 4.1.1:  it may be worth mentioning that, since DRE depends strongly also on SZA, changes in the time/seasonal distribution of the occurred changes may have affected the results.