General comments

This paper presents a validation exercise of McClear across Sub-Saharian Africa and the Maldives Archipelago. The study includes observations from several stations that have never been used, or just occasionally, for similar studies, and thus provide valuable information for users and developers of McClear.

The paper is too long. At times, it is difficult to read because it is profuse in details. For instance, it replicates in the text many data that is already in the tables. Surely, many should have been removed to make the paper more concise and shorter. It also discusses aspects of the study that could have been omitted. The language is correct and the scientific quality is good.

As a general comment, I would like to add that McClear is not run typically as an usual model that is decoupled from the inputs that are used to run it. Instead, it is normally used as a web service in combination with inputs from CAMS, despite it could also be run with other input sources. This is an important fact because, indeed, talking about “validation fo the McClear estimates” normally hides the subtlety that the validation is of McClear + CAMS. Hence, I have always thought that a better naming convention would be McClear Service to clearly state that this is the modeling approach taken, and not other.

I wonder why you chose to validate global and direct irradiance, but not diffuse. Although, admittedly, global and direct irradiances are likely more important for practical applications, the validation of diffuse irradiance is also important from the point of view of model development. The best models parameterize direct and diffuse irradiances independently and evaluate global from them. Hence, the importance of validating diffuse irradiance.

One final general comment is related with the cloud screening algorithm used here. Nothing against it, but just against the claim that it provides more confident values than the Long and Ackerman algorithm. Based on my own experience, I see difficult that the simple Lefevre et al approach can cope with the milliard of different sky situations that may hide clear skies, specially in a region like the one considered here. And I don’t know either if the Long and Ackerman approach can do it, at least, without a previous calibration of their empirical coefficients (they were set for a totally different environment).

Specific comments

P1L20. The correlation coefficient is not very much significant or relevant in clear-sky models because the cloudless irradiance is highly determined by a deterministic signal.

P1L21. What do you mean by “correctly estimated”? What is “correctly”?

P1L23. “relative bias” respect to what?

P2L38. Why that precise range for the shortwave spectrum? Can you provide justification?

P2L39-40. “Other terms… incoming shortwave radiation”. Most of these terms are _not_equivalent to SSI. SSI is an irradiance, that is, density flux of energy. It is not clear to me that solar exposure, or solar insolation does precisely refer to the same concept. Please, clarify.

P2L41. “appearing to come”? There may be scattered photons that “appear to come” from the direction of the sun Would you say that such photons are contributing to direct irradiance or they contribute to diffuse irradiance? Clearly, they are part of diffuse irradiance simply because they have been scattered. What defines direct and diffuse are the extinction processes in the atmosphere.

P2L47. “It depends on… these variables define the solar radiation impinging on a horizontal surface…” I think this is a tautology.

P7L179. Can you describe exactly the meaning of “95 % probability”?

P8L243-247. “Sea salt and dust… in and below the clouds” Is it necessary to add these comments here? They refer to the treatment of aerosols in CAMS, and it may be obscure for some readers without better context and clarification.

P9L248. “resampled in time” How? I presume it is not the same when one goes from 3 hours to 1-min time steps, than when going from 3 hours to daily time steps, for instance.

P9L255-257. “If not provided, … cell is taken into account” Specifically, how is the elevation difference accounted for?

P9L257-258. “The yearly average… total solar irradiance noted E_TSI” This sentence appears to confuse the concepts of total solar irradiance and solar constant. I suggest to review these papers: https://doi.org/10.1016/j.solener.2018.04.001, and https://doi.org/10.1016/j.solener.2018.04.067

P10L286-287. “...the results of their algorithm provide less low values of SSI… offers more confidence...” Do you think this is a true argument to assign more confidence? Then, it is easy to create a cloud screening algorithm that offers more confidence than that of Lefevre et al: simply retain even less low values of SSI.

P10L291. Wrong reference Ineichen and Perez (1999). The correct one is Perez et al, 1990: Making full use of the clearness index for parameterizing hourly insolation conditions. Solar Energy, Vol. 45, No. 2, pp. 111-114.

P10L300. “… the Rayleigh atmosphere” Strictly speaking, the scale height is rather obtained assuming a hydrostatic atmosphere at 288 K. That is, I might have a Rayleigh atmosphere (i.e., only molecules) that is not hydrostatic. Then, that height scale figure would not be theoretically correct.

P10L302. “atmospheric transmission when the reflection of the ground is null” Not clear. As defined in Eq (9), KT includes reflections from the ground.

P14L385. What is the added value of validating KT and KT_BN provided that G and B_N are being also validated?

P14L397. “Knowing that the relative standard deviation is half the relative uncertainty” Why? Can you elaborate more on this?

P28L635. “[240, 4606] nm” Why that upper limit? The lower bound is arguably set by ozone absorption. However, is there a strict limit for the shortwave spectral range?

P28L654-658. I don’t think you can get a definite conclusion out of it. It can all be just a coincidence that results from the combination of multiple sources of errors that you do not have under control.

P28L669. “The variance should also be underestimated”. I try to understand this sentence, but I am not sure if I did. Apparently you mean that adding DNI, which has some variability, and circumsolar, which also has variability, should result in a signal with higher variability than the two components separately. Hence, if circumsolar is neglected, you will underestimate the variability. However, it does not have to be like this necessarily. What happens is that you are neglecting the correlation between DNI and circumsolar and, indeed, they probably are very much anti-correlated because, for instance, an increase of AOD will reduce DNI, but most often will increase circumsolar.

Figure 9. The y-axis limits must be adjusted to the range of values that are plotted. As it is now, the figure is totally useless.

Technical corrections

P1L18. “The global irradiance, G, and…” instead of “The global irradiances G and…”

P1L25. “...the mean of B_N.” instead of “...the means of B_N”

P8L231. What is “katoandwandji”? Is it a typo?

P9L270 Is it “v1/v2” a typo?

P10L287. Awkward use of “wrote”

P32L779. “narrow” instead of “narrow to narrow”?