The paper cannot be published as is, and major revisions are required. I do not mean that the main outcomes of the work (trends of ozone and aerosol optical depths) are wrong, but several issues should be settled, or at least discussed, before the paper can be accepted.
1/ The determination of the ETC appears as the main problem, along with the estimate of its uncertainty:
1a/ the uncertainty of the mean, better taking into account the number of ETC estimates, should be used instead of the standard deviation of the ETC estimates. Otherwise, as said in the paper, the uncertainty can be lower for a time interval with few ETC estimates than for one with several estimates;
1b/ in his reply to my review, the author states that "the values of the uncertainties of the annual averages do not enter the calculation of the uncertainty of the linear trend". Instead, I think that the uncertainties should enter the calculation of both the value and the uncertainty of the linear trend;
1c/ minimum data coverage thresholds (e.g., minimum measurement days per month, minimum measurement months per year) should be fixed before averaging the data (p. 11 l. 9-15), especially in case of large gaps (due to maintenance of the instrument or clouds) and presence of a seasonal cycle. Maybe this is not an issue, but it should be specified how long is the longest period without data, if any;
1d/ I could have missed this information, but I do not understand why "an advantage of the LPM is that it takes into account the change of ETCs during the intercalibration period" (p. 16 l. 19-20). From Fig. 4, I can understand that the ETCs estimated by the LPM are averaged and fixed for an intercomparison period, aren't they?
1e/ is the Langley plot performed by correcting S_lambda (Eq. 3) for O3 absorption or by including O3 in tau_lambda? In the first case, why the ozone cycle is mentioned as a source of error (p. 5)? In the second case, how is mu_w calculated (including ozone)?
1f/ since the BSM method is basically a calibration transfer from a reference instrument, the author should explain how Brewer #017 was calibrated;
2/ the AOD differences between the LPM and the BSM methods can be easily attributed to the ETC difference (AOD difference = 1/mu * log(ETC1/ETC2)). Therefore, the deviations between the two methods are mainly not "random" (contrary to what is written at p. 15 l. 14 and at p. 16 l. 12); the observed variability between summer and winter (p. 15 l. 21-23) is an obvious consequence of air mass variation during the year; the AOD differences, always positive in the first period and always negative in the second period, are dictated by the difference in ETC shown in Fig. 5. I would thus recommend to plot the AOD differences as a function of air mass to check that the ETC difference is the main cause of deviation;
3/ as I mentioned in my previous review, an additional source of errors can be the missing SL correction in the BSM, which is not mentioned among the sources listed at p. 16. However, according to the maintainer of the Brewer operating software: "Using the real-time AOD (or even ozone) data from the operating software for anything other than a rough idea about the AOD doesn't make any sense since no corrections are applied in real time for the SL ratio changes. It's likely worse than real time data for ozone due to potential changes in absolute sensitivity in Brewers (that do not affect ozone)" (V. Savastiouk, personal communication, 2017). This issue should be discussed and its potential impact estimated;
4/ even though it is probably a minor issue (especially at low air masses), the spectral stray light effect is inaccurately taken into account, in my opinion, for two reasons:
4a/ both the slits employed for UV scans and the grating positions are different from those of DS measurements, therefore the ratio between the irradiance at longer and shorter wavelengths in the "scanning mode" may not be representative of stray light in "slitmask" mode (DS measurements). The cited study of Arola & Koskela (2004), using this method, only aimed at a rough estimate of the effect;
4b/ the assumption of spectrally constant stray light is simplistic, as pointed out in Garane et al. (2006), and will lead to some uncertainty.
Therefore, if the author wants to keep the analysis of wavelengths shorter than 320 nm, they should point out that they could be affected by the stray light, rather than trying to correct them and jump to the conclusion that the Angstrom coefficient is negative;
5/ most of all, language needs major revisions, as well as the structure of the paper:
5a/ check the whole text for the correct use of articles "a" and "the";
5b/ remove all technical details that are not fundamental for the paper (e.g., discussion about wavelengths, p. 4) and cite some reference where the information can be found elsewhere (e.g., instead of the data reduction explained at p. 8);
5c/ split the theory ("Methods") from the results. E.g., p. 11 l. 19-26 are theory, as well as p. 12 l. 5-14;
5d/ use letters, e.g. (a), (b), etc., not "right/left", to identify the subfigures in the panels, according to the ACP guidelines (https://www.atmospheric-chemistry-and-physics.net/for_authors/manuscript_preparation.html).
- replace "Operational Brewer Program" with "Brewer Operating Software";
- better use "AERONET" instead of "Cimel", since the data are provided by AERONET;
- p. 1: use the long form "spectral stray light" instead of just "stray light" at the first occurrence, to differentiate it from other sources of stray light;
- p. 3: latitude °N and longitude °E (NW/EW?);
- p. 3: order the pollution sources by relevance;
- p. 4: unless very old instruments are considered, UV, SO2 and O3 estimates are performed by all instruments. NO2 instruments are performed only by MKIV Brewers, which are no longer manufactured;
- p. 5: it is said that LPM is not recommended for low latitude stations, however Poprad-Gánovce (706 m a.s.l.) cannot be considered a high altitude station. Please, clarify this point;
- p. 5: how were the air mass thresholds of 3 (line 12) and 4 (line 32) established?
- p. 5: explain better the "polarization effect" and use the form "internal polarization effect";
- p. 5: "corrections published in Cede at al. (2006)" -> which one of the methods described in the paper was used?
- p. 6: include the physical units for the Loschmidt constant;
- p. 7: remove repetitions in lines 1-3;
- p. 7: explain why the method works best at lower latitudes;
- p. 7: "implemented", not "developed" by Vladimir Savastiouk;
- p. 7: are the alternative Komhyr (1980) and Kasten and Young (1989) formulations used for O3 as well?
- p. 8: why was the stray light correction applied at that stage?
- p. 8: explain if ND filter spectral transmittance was taken into account for AOD calculations;
- p. 8: what are the "practical purposes"?
- p. 9: the conditions 1-3 are related to the single measurement, while conditions 4-9 concern the whole day. They should be splitted;
- p. 11: "more relevant" compared to what?
- p. 11: include a formula for the correction factor;
- p. 14: what criterium was followed to define "similar" the two ln(ETC)'s, sometimes deviating by 0.1 (l. 19)? Similarly, how can you quantitatively tell that the agreement is "excellent" (p. 15) and "high level"? Compare your results to the ones obtained in a recent paper about Brewer AOD in the UV range (of López-Solano et al., 2017, https://www.atmos-chem-phys-discuss.net/acp-2017-1003/);
- p. 16: "it is not known that the BSM would take into account a change in the distance of Earth from the Sun". Does it mean that the Brewer Operating Software does not take into account the Earth-Sun distance or that you do not know whether it takes this factor into account? In the latter case, examine the source code or ask to the maintainer of the Brewer operating software!
- p. 18: since only direct sun measurements are considered in the paper, only report the ozone trend for DS;
- p. 21: explain what agricultural activities can determine the increase of AOD (wildfires?).