Referee comment on “Evaluation of the hyperspectral radiometer (HSR1) at the ARM SGP site” by Balmes et al.

The study summarizes the results of a comparison of spectroradiometer measurements over a two-month period. Data from an instrument capable of measuring total and diffuse spectral irradiance separately are compared with several other instruments in terms of total and diffuse irradiance, diffuse fraction, and derived AOD.

General Comments

The authors provide a comprehensive overview of statistical measures used to describe the correlations and regressions between the different data sets. Such intercomparisons as part of an instrument study are important to document the suitability of new instruments. However, this work is limited to the mere presentation of statistical comparative figures without deeper discussion and investigation of the causes. For this reason, I cannot recommend the manuscript in its current form for publication in ATM, but would encourage the authors to submit a new manuscript that is less descriptive. Below I give comments and some suggestions to improve and expand the content.

Specific Comments

1. It is not clear what the new technical specifications of the HSR1 are compared to the instrumentation described in Wood et al. (2017) and Nogren et al. (2022)? Is it the same instrument? The instrumental design of the HSR1 is not fully described by the authors. What kind of spectrometers are used?
2. The study lacks a detailed uncertainty analysis. The authors provide little information on calibration issues and only mention the “dome lensing effect” in the discussion section without explaining it. They should definitely expand this section. What are the single measurement / calibration uncertainties and how do they contribute to the different products?
3. The presentation of the different instruments in Section 2 should also include the instrumental uncertainty. The subsubsections (Sec. 2.2) that inform about the other instruments are quite short. Consider summarizing them without subdividing them further and showing a table with the main specifications.
4. P2L46: “radiometer with no moving parts is now available called the hyperspectral radiometer (HSR1)” – I would leave out the phrase "no moving parts". It would be better to say that no rotating shadow band is required.
5. P2L53-L59: I had difficulty understanding the brief description of how to separate total and diffuse irradiance. Only the references given made it clearer. Even though it is redundant to the other publications, I would recommend a sketch for better understanding. Since these are instrumental details, I would place this information in Sec. 2.
6. P2L53: “the shadow pattern allows one of the seven sensors to be illuminated unobstructed by the shadow pattern, which measures the total irradiance” – According to Wood et al. (2017) it should be I_max+I_min, which gives the total irradiance. Please clarify.
7. P3L79-L83 + Fig. 1: I think that this example plot is not needed at this point. Data coverage is reported at the beginning of this section. All the detailed information about the reasons for the downtime may be less important to the reader. Try to shorten them. Perhaps show a time series of the radiation data along with the cloud cover data.
8. The second part of Sec. 2.1 should contain more technical details of the HSR1 instrument. Here the spectra from Fig. 1 would fit.
9. P5L98-L103: The spectral range limitation could be better justified by using radiative transfer simulations that show the low performance at the edges of the spectral range. Is it really stray light that is causing the low performance? Have you done lab tests with edge filters to analyze this?
10. P5L108: “lensing effect” might be important – Please elaborate. It is referred to Sec. 5, but there is no deeper discussion.
11. Section 2.2: Please give uncertainties for all instruments / products.
12. P6L143: What is the wavelength range that is covered by the instrument?
13. P6 Sec.2.2.3: Too many details on instrumental failures. It is sufficient to say, that only cloudless conditions could be considered due to instrumental issues.
14. P7L150: I am not sure that the comparison of PAR data is really necessary in this study, since it is just another quantity based on spectral integration.
15. 3: Please discuss the uncertainty of the AOD retrieval.
16. P8L183: How does tau_gas depend on the vertical profile of temperature and pressure? Is ozone the only type of gas that contributes?
17. The figures are well presented but include all formula signs in the text. Example Fig. 2: “F_total”.
18. All frequency histograms give the mean value of the two-month period. I am not sure if this is an appropriate measure since the data are not normally distributed and may have multiple modes.
19. 2d: Only the regression line is shown here. What does the scatter plot look like?
20. All scatter plots / frequency histogram: Is the scatter between the different instruments within the measurement uncertainties?
21. 4: same as Comment#16. To show a possible wavelength dependence, it might be helpful to plot the RMSE as a function of wavelength. Table 1 is sufficient for the interpretation of the sign of the bias.
22. Table 1: Way too many numbers. I recommend reducing the data to the HSR1 comparison only. The AOD results should be given in a separate table in Sec. 4.2.
23. 4.1.1: I don't see any gain in information when the comparison results between the other instruments are shown. It is a bit monotonous to give all the numbers in the text which can be read from the table. The same holds for the AOD comparison P18L340-L351.
24. P21L363: “The spectral AOD results at all wavelengths are similar to those at 500 nm (Fig. 6)” – rewrite. The absolute numbers are not similar.
25. 7: same as Comment 21: To show a possible wavelength dependence, it might be helpful to plot the RMSE as a function of wavelength.
26. Maybe swap Sec. 4.3.1 and Sec. 4.3.2. It would make more sense to look at the diffuse spectral components first before showing the integrated values.
27. P22L408: “The motivation of this comparison is to understand if the HSR1 integrated diffuse ratio captures the diffuse ratio in the absence of a diffuse broadband irradiance observation (e.g., only total broadband SW measurements) despite measuring only a portion of the solar spectral range.” The wording could be improved. Do you mean “broadband” in the sense of solar broadband? The spectral integration of the measured total and diffuse irradiance gives a broadband irradiance but not fully covers the solar spectral range. To identify the missing fraction could be analyzed more deeply by using a radiative transfer model.
28. P23L412: A lower mean diffuse ratio is reported for the HSR1 than derived from the Radflux instrument which covers a broader spectral range. I would expect it to be the other way around, since with increasing wavelength the diffuse ratio decreases strongly, as radiative modeling could show.
29. P23L422: “The mean diffuse flux error is -16.7 and -7.9 W m-2 for all times and clear-sky times, respectively.” Perhaps it would be better to show a distribution of the bias illustrating the different modes.
30. P23L423: “Noting the measurement uncertainty of ±3% in the diffuse flux (Michalsky and Long, 2016), only 16.5% (all times) and 18.3% (clear-sky times) of the diffuse flux errors due to considering the HSR1 diffuse ratio are within measurement uncertainty.” I do not understand this sentence. Please rephrase.
31. P25L471: “The SASHe clear-sky spectral diffuse ratios were also compared at 415, 615, 673, and 870 nm. The results are found to be similar to the 500 nm results.” – rephrase. The absolute numbers are different.
32. Discussion section: It is not really a discussion of the results. Definitely more content and deeper thinking about the reasons, uncertainties, and relationships between the quantities is needed here. The information on calibration and post-processing is quite vague.
33. Summary section: This section is a way to long. It repeats all the numbers which is kind of exhausting for the reader. Try to reduce it to the main points.
34. Conclusion section: I do not find any conclusion here. What can the reader learn from this study?

Technical Comments

1. P3L78: “C1 (36.607322 °N, 97.487643 °W) and E13 (36.604937 °N, 97.485561 °W).” – Give the distance.
2. P7L158: “Ozone satellite”: rephrase
3. P7L151: “PQS1”: What does it mean?
4. 4: Symbols and labels are too tiny.
5. P15L247: “MFRSR filter” à MFRSR narrowband filter
6. 8c: No y-axis label.
7. P23L418: Here and at other locations the term “flux” is used. Keep using the term irradiance.
8. P24L437: “F” irradiance in italic