GENERAL COMMENTS

The manuscript by Zhao et al. analyzes the atmospheric conditions affecting the primary calibration of reference Brewers at four sites, with particular reference to the short-term variations of the total ozone column. A modelling framework is developed to simulate the effect of the ozone variations on the extraterrestrial calibration factor, using MERRA-2 reanalyses as input. The study aims at assessing: "(1) why Brewer primary calibration work can only be performed at certain sites... and (2) what is needed to assure the equivalence of calibration quality from different sites" (lines 104-106). While the answer to the first question is rather obvious, the second research question is very relevant to the ozone science. The paper is generally written in a clear way. Based on these considerations, I would recommend publication of the manuscript after some corrections.

SPECIFIC COMMENTS

• Structure. It may be a matter of taste, but I would recommendation to follow a more traditional paper structure better highlighting the methods, the results and the discussion. I feel a bit confused, for example, when reading that Sect. 3 is titled "Primary calibration" and Sect. 3.2 is titled "Primary calibration method": what difference should the reader expect from the two sections?
• Different effects of short-term ozone variability. I am sure that the authors can introduce this topic in a more clear and tidy way (lines 320-330). Indeed, daily variations in ozone and instrumental factors can result in both random "noise" in the ETC determinations (appropriately tackled with type-A evaluation of the uncertainty) and systematic effects, notably in presence of recurring daily or sub-daily patterns (e.g., photochemistry? local pollution? instrumental artifacts?). I think that this distinction requires further discussion, especially if the authors examine the linear and parabolic terms of the daily variations (e.g., how regularly these patterns occur?).
• Relation between simulated and observed ETCs and reliability of MERRA-2 reanalysis. The authors state that:

1. the "large day-to-day difference is mainly due to short-term ozone variability" (l. 297)
2. they use MERRA-2 to "isolate the short-term ozone variability impacts" on ETCs (l. 360 - what does "isolate" mean, exactly?)
3. "the impact of linear and quadratic ozone variations can be independently assessed" (l. 405)
4. the uncertainty of MERRA-2 reanalysis is still too high to be used for correcting the single ETCs for short-term ozone changes (lines 600-603).

All things considered, a straight question is: is there a direct correlation between the ozone variations from MERRA-2 (e.g., in terms of the "b" or "c" coefficient) and the individual observed ETCs? Or should simulations using MERRA-2 data (and their good agreement with the obseved behaviour, e.g. in Figs. 5e-h) only be interpreted as an average indication of the effect and its magnitude, depending on the site and the considered season? Can the authors further elaborate on the results reported at lines 406-408?

• Sect. 4: is SO2 an interfering factor at MLO?

TECNICAL REMARKS

• Title: I'm not a native English speaker, but "site-specified" sounds a bit odd (should it be "site-specific"?)
• l. 29, "stable short-term ozone field": "stable ozone field in the short term"
• l. 36: notice that if R6 units are divided by the differential absorption coefficient, the same quantity can be expressed in DU (relative to airmass 1). The equivalent quantity in DU can thus be specified in the text for greater clarity
• l. 117: it should be mentioned that the second "channel" is used for SO2 retrievals
• l. 124-125: this formula should be written in a distinct line, so that the definition of F can be referenced more easily if needed (e.g., line 225)
• l. 125-126, "the last four longer" --> "the four longest"
• l. 146: notice that Cede et al., 2006 employ MkIII Brewer, which clashes with the premise "NO2, by Mark IV only"
• l. 168: "wavelength" --> "wavelengths"
• l. 191-195: isn't it a repetition of what is already said in the Introduction?
• Eq. (2): notice that the equation is different from (1). In addition to the 10 factor, the sign of the ozone term is opposite. Please, use only one convention for the sign of the differential coefficient (e.g., define it positive). Also, use either the "F0" or "ETC" expression
• l. 237: mention that std(O3) requires a first guess of ETC
• l. 262-264: unclear. ND filters can indeed have some color
• Eq. (4): please add the j subscript below the sum symbol
• l. 282: are data from Eq. (3) corrected by the ND filter effect, as mentioned at lines 277-278? Otherwise, differences can arise from that effect, too
• l. 291, "corresponding to ... slant column ozone": ... times the differential absorption coefficient
• l. 291-293: from my understanding, differences do not arise from the equation used, but rather from the treatment of the ND effect
• l. 292: "Fig. 1" --> "Fig. 1a"
• l. 294: "Fig. 2b" --> "Fig. 1b"
• l. 327: please explain why low-latitude sites have less ozone variability, for the unexperienced reader
• l. 331: why should it be more "consistent" to use eq. 2 instead of Eqs. 3-4, which were already employed for Fig. 1?
• l. 334: I might have missed this information, but are the measurement schedules similar? How does the number of direct-sun measurements scheduled in a day by the 11 Brewers affect the number of successful Langley events?
• l. 337-339: this information should be reported in the corresponding figure, e.g. as column titles
• Eq. 6: what "t"'s are used? Are they the exact time of measurements for the different instruments?
• l. 407-408: this is a bit counter-intuitive, as correlation with airmass should be more "dangerous", as found by other authors (e.g., doi: 10.1029/2006JD007248). Also, can full-day Langley plots, instead of half-a-day events, be used to detect and/or fit the linear daily trends (e.g., doi: 10.5194/amt-7-4009-2014)? Can the authors elaborate more on this?
• l. 411-412: comparison of the two figures reveals qualitative/average agreement between simulations and real observations. It would be more interesting to have individual ETCs plotted in the same figure, to check whether there is 1-to-1 correspondence
• l. 413, "show the" --> "show that the"
• l. 414, "simply due to this short-term ozone variability": can the authors plot these variations vs the "b" or "c" coefficients calculated for each simulated Langley event?
• l. 438: explain the reason of the stronger short-term ozone variability at midlatitudes in wintertime, for the unexperienced reader
• l. 488-489, "as a result of air pollution": if this is the case, should the "week-end effect" be visible on the ETCs?
• l. 500-501: please, explain the calculation relating AOD = 1 to ozone uncertainty = 1 DU
• l. 505-509: is the amount of aerosol (AOD) the only affecting factor, or is its variability (e.g., systematic cycles) during the day of any relevance, too?
• l. 587-588: as this is a recommendation, who are you referring to? Should the Brewer operators or processing centers perform this check? Can this be implemented in some publicly available software?
• l. 636-637: "individual hourly values", of which series?

General comments:

The manuscript "The site-specified primary calibration conditions for the Brewer spectrophotometer" by Xiaoyi Zhao et al., aimed to answer two primary questions related to calibration of the Brewer spectrophotometers. The first question addresses site-specific factors, i.e., why the calibration procedure will not work well at certain locations.  The second question, closely-related to the first question, deals with the required conditions to achieve a certain calibration quality.

The answers to these questions are important to the assessment of the measurement quality of the ground-based Brewer (and Dobson) measurement networks, and therefore important to Ozone research.  

The use of available auxiliary data and the development of a modelling framework using MERRA-2 to answer these questions are done in an innovative and convincing way.   

The manuscript is well written and well presented. The methods employed are scientifically robust and they are clearly explained. The figures are clearly illustrated, except for one or two that can be easily improved. This manuscript fits within the scope of AMT. Therefore, I recommend its publication after addressing the comments of Reviewer #1 and some of the minor comments below.

Specific comments:

1.    The Authors briefly mentioned the ETCs from the ICF (Instrument Constant File). I think it should be clarified in a few words what this is, and how the ETC values in this file are obtained. 

2.    Related to comment #1 above, it is my understanding that Brewer spectrophotometers have internal quartz-halogen lamps, which are used for the purpose of monitoring instrument stability and changes in ETC values. It would have been interesting to know the results from these regular lamp tests and how the lamp test results can be employed to support the calibration method. I would have preferred if there was a discussion about it.

 

3.    “short-term ozone field”: It took me a while to decipher its meaning, perhaps it is a modelling term. I would recommend clarifying it when the term first appears on page 1.

     

Technical Corrections and Suggestions:

As Reviewer #1 already mentioned, “site-specific” is more appropriate than “site-specified”, which appears in many places in the manuscript. 

P.5, Line 142: “are can be found” -> are found OR can be found 

Fig. 2. Among the nice figures, this is probably the only one I find difficult to decipher. It is an important figure, perhaps this could be improved. The panels are too small. Also, I would suggest to avoid using red and green markers in the same figure. 

P. 13, Lines 347-348: This (second to the last) sentence needs to be revised or rephrased. 

P. 23, Line 599: “… the Brewer spectrophotometer taking the most accurate TCO observations among ground-based instruments …”. This claim seems like the Brewer has been validated against a “true” measurement of TCO, and then compared with other ground-based instruments, which are also validated against a “true” TCO. If so, please provide references. 

References: Multiple papers of some authors are not in chronological order, e.g. papers by Kerr et al.