General Comments:

This article presents a new data set of UV index observations for an area of the world that generally lacks observations.  A comparison is presented with UV indices derived from various satellite products.  On the whole, these items are worthy of publication in some venue.  However, based on the provided review guidelines as I understood them, I could not confirm that the submission meets the objectives of a data description article for the Earth System Science Data journal.  Quoting from the aims and scopes of the ESSD (https://www.earth-system-science-data.net/about/aims_and_scope.html),

“Articles in the data section may pertain to the planning, instrumentation, and execution of experiments or collection of data. Any interpretation of data is outside the scope of regular articles. Articles on methods describe nontrivial statistical and other methods employed (e.g. to filter, normalize, or convert raw data to primary published data) as well as nontrivial instrumentation or operational methods. Any comparison to other methods is beyond the scope of regular articles.”

This article does describe how the data was gathered and processed, including calibration information, but this is actually a minor portion of the text and mostly references other sources.  The bulk of article describes interpretation of the data, for example the presence of cloud enhancement, and comparisons of radiometer data with results determined by other methods (satellite and model-based products).   The short summary states that the article concerns the validation of a new dataset from the measurement of ultraviolet radiation, but the validation described in the lines 110-115 is actually of the satellite and model based numerical products given the ground-based radiometric measurements.

Moreover, a stated criterium of a ESSD contribution is that it should be easy for any interested person to use the associated archived data, but the data available at https://zenodo.org/record/4572026#.YH8SL2gpC9Y doesn’t quite meet that standard.  The meta data should provide more information, including what zone is used for time (UTC? LT?), what is a UV index, and what is meant by an “Instant Reading”.  In one of the data sets, AnseQuitor, the same observation occurs in triplicate at each time point which appears to be an error.  The files include data at “night” (SZA >90 but not reported) which are all zero and makes the files unnecessarily long (zero data is omitted for the St-Denis data).  The cloud fraction data includes the red-blue ratio the derivation of which is not described in the article.  Finally, the data doi only includes the radiometer (Kipp Zonen erythemal) data and not the other data referenced in the paper, including the Bentham spectroradiometer data, TOC and UVI derived from the satellite and model products.

Calibration procedures are mentioned in the article but details are omitted, in particular when calibrations were performed for the data presented.  The article does mention plans for future operation and procedures for future calibrations.  This information will be of interest to the reader but is not directly relevant to the archived data.  The organization would be improved by segregating such information in a separate section on titled something like “future operations”.

This is the first paper I have reviewed for ESSD and perhaps I have applied the criteria for publication too narrowly.  For this reason, I have not recommended outright rejection and leave it to the editor to make the final judgement.  Whatever course is taken,  the mss and data sets with revisions should be appropriate for publication in this or some other venue for reporting geophysical results.  In the revision, I recommend attending to the points made above and a number of specific issues as follows:

3.1 Calibration – This section refers to results on Radiometer ADs and RDs relative to the Bentham spectroradiometer but only qualitative description is given.  Comparing this section and the data section lines 95-109, I am confused about what calibrations were used and how they were done.  Line 97 says that the radiometers will be calibrated every two years.  How about the calibrations for the presented data?  In section 3.1, differences are described between radiometer based and spectroradiometer based UVI for the “recent recalibration”.   What was the calibration of the radiometers for this comparison?  Later in the paragraph, it is stated that “differences were used to recalibrate the radiometer”.  But isn’t that what is usually done in a calibration against a reference instrument?

What are the criteria for determining which measurement are “clear-sky” and what is meant by a “clear sky day”  Please specify.  The text says, e.g., that there are 16% “clear sky days” at Antananarivo, but the data set AntananarivoSKYCAMVISION has < 1% of 30s resolution data with CF=0.

Line 93 “on station Juan de Nova”

Should read “one station at Juan de Nova”?

Line 95 “All stations are now equipped with a Kipp & Zonen UVS-E-T broadband radiometer.”

Table 3 and later text states that some of the radiometers are the SUV-E model

Line 96 “The raw UV measurements obtained by the radiometers are reprocessed considering the calibrations and TOC measured simultaneously.”

Please describe and/or give reference for the reprocessing procedure.  The reference given in Table 2 describes how the KZ radiometer was calibrated in Davos, but what is the current procedure?

Line 101  “on a smaller mesh size”

Not clear what this means

Line 212 “Table 4 presents the different radiometers and their current locations, along with the date of the next calibration”

This information is in Table 3

Line 225 Figure 2 shows the period covered by each data set at these four stations.

The figure only has one line covering data for the broad-band radiometers, does not show each station separately

Line 315 – UV enhancement by cloud scattering is a well known and widely occuring phenomenom, would be appropriate to cite the Sabburg and Wong (2000) paper here, also see Badosa et al. (2014)

Line 347 “The density of the corresponding data set for each month of the year is represented in Figure 7b”

Please define what is density in this context, proportion of what?

Line 394 - The correlation coefficient between the satellite or model estimates and the ground-based measurements was greater than 0.9 at all stations except Mahé and for all datasets except OMUVBG.

Revise to state that correlation applies only for clear sky conditions

6-Conclusions

Perhaps the authors can comment on why of the three stations at similar south latitude, Antananarivo, Anse Quitor,  and St-Denis,  the UV-index, even for clear sky, seems to be systematically lower at St-Denis.  For example, the noon mean CS UVI is ~10 at St.-Denis but ~12 at Anse Quitor.  The maximum CS UVI is ~18 at Anse Quitor and ~17 at Antananarivo, but <15 at St-Denis.

Figure 1,  Map – add Latitude and Longitude

Figure 3 caption– What do the gray bars represent in the histogram sub figures?  This also applies to the Appendix figures

Figure 5. The caption reads “Diurnal Cycle of UVI at ST-DENIS.”

The figure actually has all stations, not just ST-DENIS.

Figure 7 & 8 captions – Describe the data density subplots (b) and (d)

Figure 7c and 8c – The annual mean difference CS-AS UVI seems to follow very closely the DJF mean – Is this correct?  Seems like the mean should be approximately in the middle of all the monthly means, but this is not the case for much of the plot, especially in 8c.

Table 1 – Region location for Juan de Nova station should be Ile Juan de Nova

Reference missing bibliographic information: Pastel, M., Pommereau, J.-P., Goutail, F., Richter, A., Pazmino, A., Ionov, D. V., and Portafaix, T.: Construction of merged satellite total O3 and NO2 time series in the tropics for trend studies and evaluation by comparison to NDACC SAOZ measurements, 2014.

References:

Badosa, J., Calbó, J., Mckenzie, R., Liley, B., González, J.‐A., Forgan, B. and Long, C.N. (2014), Two Methods for Retrieving UV Index for All Cloud Conditions from Sky Imager Products or Total SW Radiation Measurements. Photochem Photobiol, 90: 941-951. https://doi.org/10.1111/php.12272

A useful manuscript detailing UV measurements in a part of the world that has, until now, had limited coverage. I expect it should find a suitable home somewhere, but as Reviewer #1 notes, it does appear to fall slightly outside the remit of EESD by a strict reading of the journals aims and scope. This could be addressed by a shift in emphasis to providing more details about the ground instruments and their capabilities and calibrations, and a summary of the resulting data, and less emphasis on the intercomparison to satellite datasets. However that specific point I leave for the editor.

Aside from the issue of whether the manuscript properly fits within EESD's remit, I think the manuscript would anyway be improved by this shift in emphasis and some restructuring. There would be much to be gained by adding details of the calibration procedures, more details on the sites and their localities, more details about what instruments are running when at each station, more details about the data processing. This would enable the reader to assess the quality of the resulting data, which is not really possible at present. The comparison to satellite data tells us more about the disparities between different satellite and model products than it does about the ground data — ground data usually being used to validate satellites, rather than the other way round. I found the two sections comparing ground based data to satellite data less helpful, perhaps because the thrust is that there are differences between the Bentham and the satellites due to their different footprints and processing techniques, but little effort to resolve these differences. As a result it tells us little about the Bentham and UVR data. More details and depth on the calibration procedure and instrument characterisation on the other hand would assist here. The conclusions are also heavily biased to interpreting the satellite aspect rather than the novel ground data.

In general I would recommend the authors restructure the paper. The introduction is mostly fine. Then perhaps something as follows:

- site selection (how were they chosen)

- site details (latitude, longitude, altitude, urban, rural, general surrounding habitat, other ancillary measurements e.g. SAOZ)

- instruments and which at which sites

- instrument maintenance protocols and frequencies, characterisations (cosine error, spectral responses), calibrations and data processing (include levels of data)

- data products and pathways (leading on from data processing)

- presentation of summary data: diurnal variation of UVR and CF at each site, monthly means / annual variation, depending on length of measurement period

I think most of this information will be available to the authors if it is not already in the manuscript, but, for example, putting all the information about calibrations together will help (at present it is spread between sections 2.1 and 3.1).

Some more specific points follow:

ll 1-2 Rephrase to make it clear that the spectroradiometer is a single instrument shared across the network, rather than each site having a spectroradiometer. Likewise for the all sky cameras that not all sites benefit from such an instrument.

l 2 "Homogenously" is not quite correct, there are not an equal number in each country; perhaps "spatially" or "geographically" would be better. How these sites were selected would also be interesting to include.

l 19 "increases more than 4". I presume this does not mean a factor of 4, but it should be made clearer to the reader.

l 48 Insert comma after "radiation" and remove after "conditions"

l 82 Region should not have a capital here

ll 92-95 This would be better changed to "all sites have operational UVR instruments (...), and four stations (...) have operational all sky cameras" or similar 

l 96 Which model spectroradiometer?

l 97 How are the TOC and calibration measured simultaneously? Or do the authors mean that the calibration is dependent on TOC (and also SZA?) and applied after measurement? How was the calibration obtained? How often is it reassessed?

ll 98-100 What is the calibration frequency so far? This is more important than an anticipated future 2 years cycle for the data presented. How is the cross-calibration carried out? How is the Bentham calibrated and what are the details of its calibration / traceablity? Why was the Saint-Denis calibration method changed from the Davos one to the Bentham-derived one? How does this affect the data? Do the authors mean PMOD when they state "Davos"? If so, this should be stated more clearly and acknowledged.

l 101 What do the authors mean by mesh-size (graduations in TOC in the lookup table for calibration)? If so, how does this relate to the cost of the instrument?

l 102 How much difference is there between OMI TOC and SAOZ TOC at the stations with both, and how much does this affect the UVR calibration?

l 105 It would be useful to know what level 0 and level 1 data is as well.

l 107 It would be useful to provide at least brief details of the cloud segmentation algorithm and any known issues or bias in addition to the citation to assist the reader.

l 96-197 Using the abbreviations AD and RD makes reading the following section much harder. Suggest they be removed and spelt out.

l 208 Remove space before period.

l 211 BENTHAM should not be fully capitalised here

ll 231-234 This detail should be left for the figure caption.

l 241 Absolute and Relative Differences should not be capitalised

ll 248-249 Yes UVI-RADIO and UVI-BENTHAM should be aligned, but as the authors are presenting the data, and the quality of the data depends on the quality of the calibration, how well this has been transferred between the two instruments would be very interesting to include.

l 252 M-RD: these designations make reading the text harder

l 289 Likely smaller than 100km as well, but including some details of the sites' local environment would complement this point

l 300 How long is the measurement period? Please provide details.

l 317 Enhancement suggests a ratio (or factor), but as before do the authors mean a difference of 2 to 3?

l 328 How long is the data set being discussed?

l 369 Rearrange sentence so does not start with "10 am"

ll 380-390 This information would be better in a "data availability" section

Fig 1: It would be useful to distinguish the stations, perhaps by colour, according to the instruments that monitor at each site.

Fig 2: This mainly shows satellite timelines. It would be better to show the timelines of the UVR instruments which give most of the geographic coverage and the all sky cameras which are the focus of the manuscript

Fig 5: Is this UVR data or Bentham data?

Fig 6: I am not sure whether the length of the data series is sufficient to call these "climatologies"