Review of

Comparison of total ozone measurements in Melbourne, Australia, performed with a low-cost micro spectrometer and a Brewer MK-III

Kåre Edvardsen, Matt Tully, Steve Rhodes

Submitted to https://doi.org/10.5194/egusphere-2025-87 Preprint. Discussion started: 17 February 2025

General comments:

This paper provides a description of the methods to derive total column ozone (TCO) from a very low-cost instrument measuring global ultra-violet irradiance. The paper is well written, well structured and the methods are clearly described. It is recommended for publication after addressing the comments below.

The paper focuses rather on the method of TCO retrial rather than a sound analysis of the performance of TCO compared to a reference instrument. Therefore, the general comments are divided into two following parts:

Method:

Commonly, TCO is derived best by direct sun measurements. The paper uses global UV measurements, which implicate parameters of uncertainty such as clouds in front of the sun, clouds in the sky, effective albedo from the surrounding terrain and cosine response of the diffusor. All these effects would not occur when using direct sun measurements. The authors argue that in many places worldwide cloud coverage hinder direct sun measurement for TCO. Are there so many places with such conditions? Maybe the performance of the low-cost instrument would be much better when using direct sun measurements. Has this ever been tested with the instrument?

The method includes many parametrizations based on empirical functions. I have lost the overview. Maybe the authors can provide a summary of all the empirical parametrizations, the number of measurements needed for the parametrization and, if possible, an estimation of the uncertainty of the parametrization and its implication of TCO.

The parametrization also arises the question if the method is also suitable for other placed world wide, once it has “calibrated” with a Brewer at a specific site. Or must the method be “calibrated” for each location worldwide separately with a Brewer?

Performance:

It is not clear if the authors have compared daily averages with quasi synchronous measurements for STS GI and Brewer DS. I suggest comparing synchronous measurements within e.g. 10 minutes intervall and also show the individual measurements to indicate the variation of TCO.   

Commonly these small low cost spectrometers suffer from straylight, which biases TCO at low solar zenith angles or high ozone slant columns. In order to assess the performance of the instrument the comparison between the Brewer and STS GI depending on ozone slant column (airmass * TCO) should be provided.

Temperature dependency. Indeed, the dependency of the small array spectroradiometer is a crucial issue of these instruments. Ambient temperature causes wavelength shifts, linearity and detectability problems (signal to noise ratio etc.). I am surprised that a simple parametrization (Eq. 2) can account for that. It would be worthful to calculate the uncertainty of TCO based on the parametrization and its applicability for other instruments and other locations.

Furthermore, the model includes the ozone absorption cross section, which depends on the effective ozone temperature (basically stratospheric temperature). How sensitive is the algorithm and resulting TCO on effective ozone temperature?

Specific comments:

Page 2: line 32: How significant is the fraction of the price of the Brewer. To my estimate it is about 10% of a Brewer, which is already low cost.

Page 2, line 47: Is the MK II Brewer a double monochromator with insignificant stray light impact?

Page 2, line 60: Can the method be described as a cloud correction?

Page 3 line 89: The reference is missing

Page 4 line 95: What is the slit function and full width half maximum of the instrument, resulting from 25 micron slit?

Page 5 line 126: Do you mean ozone slant column (=airmass * TCO) - > see comment above.

Page 5: 132: and EQ 1 is this parametrization also applicable at other locations

Page 5, line 146. Again, what is the full width half maximum of the slit function?

Table 1: It would be worthful to indicate the performance in percent

Caption Figure 2: SZ -> ZS

Figure 4. The ratio between Brewer and STS would be more helpful

Page 11 line 280: Why is TCO increasing with temperature? Wavelength shift?:

Page 11 line 286: I suggest using quasi simultaneous measurements.

Page 12 line 310: At what conditions is STS comparable to the Brewer? What are the limitations?

Page 13: line 324. The conclusion should not end with a bullet list. Maybe a closing sentence would help.

General Comments:

The paper by Edvardsen et al. presents a newly developed low-cost total ozone instrument, including technical descriptions, ozone retrieval algorithm, and some validation. The new instrument shows good potential to be a low-cost instrument performing network-level monitoring. However, the retrieval algorithm currently developed seems to be oversimplified and some extra validation work would be good to confirm its performance. In addition, the calibration of the new instrument relies on a collocated Brewer spectrophotometer. Is the calibration at a given site transferable to any other locations (where we might not have a Brewer)? Most likely it will not be straightforward, such as it would depend on the altitude of the site, local ozone and aerosol profiles, etc. Some discussion and explanation is needed. Nevertheless, this work is a good match with AMT and within the scope of the journal. I recommend publishing this work after addressing the following issues.

Specific Comments:

Line 32: to make this claim more convincing, maybe give some rough numbers here for the listed instruments. At least, Pandora and BTS-solar instruments are commercially available.

Line 55-56: In fact, there are still lots of good improvements on Brewer DS ozone. E.g., Savastiouk et al., 2023.

Line 57-60: To justify the claim, some more quantitative description is needed. E.g., % of DS data removed due to cloud conditions.

Line 60-64: some references for Brewer ZS method and data quality are needed.

Line 84-85: The cloud effect could affect DS, ZS, and even GI TOC. I understand the latter two should have better performance than DS in cloudy conditions. Some more quantitative description is needed to support such a claim.

Line 110: Why does Brewer DS have a cut off at 63 degrees (mu value around 2.1)? Is this a single or double Brewer? For double, typically the cut off could be 75 degrees (mu around 3.5).

Line 116: is there an option to control the exposure time? It is a bit surprising that some data has to be discarded due to this.

Line 120-121: some observation SZA ranges could be provided for the reader to understand the capacity of STS. Can STS perform observations when SZA is close to 90 degrees (like ZSL-DOAS systems)?

Line 165: given this is a new effort, why use “Bass and Paur”, not anything we know has better performance (e.g., SDY, see Redondas et al., 2014; Voglmeier et al., 2024)?

Line 155-156: The figure shows the method will lose sensitivity for SZA< 40 degrees. The good news is the method would have good sensitivity for TOC from 300-500 DU, when SZA is in the range of 40 to 80 degrees. This should be good enough for most mid-latitude conditions (but could be challenging for low- and high- latitude areas). I would expect to see a detailed exam on the algorithm detection limits or uncertainty budgets.

Line 161-162: So, the LUT is only good for mid-latitude, correct?

Line 172-175: I am a bit surprised that the author directly used the ratio between measured and modelled irradiance. Given that the spectrometer is not absolute calibrated, how can we ensure such ratio close to unity? Such ratio will be different for different instruments, correct?

Line 193-196: I am concerned about this issue. First, this 5% offset per 10 degrees C is not small. For some days in mid-latitude regions, a 10-degree C daily variation is common. The temperature for the warmest and coldest days is not given. But, if we use the 44.3 degrees C from the second warmest day (reported in the paper), the correction (or bias) is alarmingly 12%. Is this from wavelength drifting? For the algorithm, is there any “dynamic” wavelength registration or is it just using values from fixed pixels? Is this STS system placed outdoors without any temperature control? In any case, the author should show the correction factor’s linear regression plot here (and also linear regressions of STS TOC vs. Brewer TOC, before and after applying this correction). Sophisticated correction is “cheaper”, only if we cannot do it correctly. I agree with the author that some level of T control is needed for this system.

Line 207-208: why STS has less data points than Brewer DS and ZS? Even including those four days of thunderstorm, STS still has fewer data points than others. Any idea? I would expect that this STS GI method should produce more data even in cloudy conditions, based on the claims in previous sections.

Line 238-251: it is a bit strange to show a trend for such a small dataset, which even could not cover a full season cycle. But, it is a bit surprising that Brewer DS and ZS ratio has a significant trend. If this trend is real and keeps increasing, some ZS calibration might be needed (i.e., the nine coefficients should be recalculated). When was the last DS and ZS calibration done for this Brewer? Could the author include scatter plots for such a comparison?  Also, I agree with the author that the poor performance of STS in the first 30 days could be due to inaccurate T correction. I am wondering if the author calculated the T correction parameters based on daily mean data or high-resolution observations. Any estimation on the uncertainty from this correction?

Line 261-267: Not just cloud but also aerosol and surface albedo could affect the results. Cloud optical depth, aerosol optical depth, cloud layer heights, aerosol layer heights, surface albedo, and ozone profiles could all affect the ratio. This simple ratio method reminds me of “color-index” commonly used in the DOAS community, which is also just a simple ratio of two wavelengths (e.g., Wagner et al., 2016; Zhao et al., 2019). By any means, to convince the reader that all the suggested factors won’t show a big impact, some detailed modelling work should be done (even better to understand the uncertainty).

Line 275-276: A typical timeserver nowadays can assure more than enough accuracy than ±20 s. The major issue for this method and instrument came from other places. This is why I would suggest some level of uncertainty in budget estimation.

Line 279-281: based on the information provided, this is not a simple electrical noise, but something systematic. My first guess would be the wavelength registration issue. Is there any compensation for wavelength drifting? The resolution of the selected spectrometer is pretty low (3 nm), but do we see any slit function changes?

Line 287-289: Simple solutions such as white painting, shading covers, and better ventilation could be done to reduce direct heating from the sun. In any case, if the system could only survive up to 40 degrees C, it could not be used in many places. Also, is the simple linear correction for T only valid within 40 °C? Note that in Line 199, it claims that the system could work from 0 to 50 °C. Please provide consistent information.

Line 310-312: some quantitative description is needed. It is a bit hard to understand to what extent STS consider the condition as “too cloudy”.

Line 313-314: give this is a very simple instrument, such stability for six months is good but not enough. Lots of total ozone monitoring instruments need to be as stable as 2 years with minimal maintenance supports (such as entrance cleaning). Also, it would be important to see if there is any seasonal bias between STS and Brewer. Is this STS no longer co-located with Brewer?

Last, in fact, Brewer also can retrieve total ozone via UV irradiance, not just DS and ZS. Some solid works were done almost 30 years ago. I would suggest the author check (Fioletov et al., 1997). For example, I would suggest using the log scale for the ratio and using all wavelengths instead of just one (312 nm). 

Technical Comments:

Line 26: left parenthesis is missing

Line 82-84: need to rewrite this sentence to make it clearer

Line 89: reference is missing

Line 91: definitions for WOUDC, NDACC, and PGN are needed.

Line 93: change “Environment & Climate Change, Canada” to “Environment and Climate Change Canada”.

Line 102: delete the repeated “in a”

Line 112: change “clods” to “clouds”

Figure 2: what are “SZ” measurements?  Please use consistent naming for the data. Also, the legend shows DS, ZS, and STS, while the caption talked about “ZS and GI measurements”.

Reference

Fioletov, V. E., Kerr, J. B., and Wardle, D. I.: The relationship between total ozone and spectral UV irradiance from Brewer observations and its use for derivation of total ozone from UV measurements, Geophys. Res. Lett., 24, 2997–3000, https://doi.org/10.1029/97GL53153, 1997.

Redondas, A., Evans, R., Stuebi, R., Köhler, U., and Weber, M.: Evaluation of the use of five laboratory-determined ozone absorption cross sections in Brewer and Dobson retrieval algorithms, Atmos. Chem. Phys., 14, 1635–1648, https://doi.org/10.5194/acp-14-1635-2014, 2014.

Savastiouk, V., Diémoz, H., and McElroy, C. T.: A physically based correction for stray light in Brewer spectrophotometer data analysis, Atmospheric Measurement Techniques, 16, 4785–4806, https://doi.org/10.5194/amt-16-4785-2023, 2023.

Voglmeier, K., Velazco, V. A., Egli, L., Gröbner, J., Redondas, A., and Steinbrecht, W.: The transition to new ozone absorption cross sections for Dobson and Brewer total ozone measurements, Atmospheric Measurement Techniques, 17, 2277–2294, https://doi.org/10.5194/amt-17-2277-2024, 2024.

Wagner, T., Beirle, S., Remmers, J., Shaiganfar, R., and Wang, Y.: Absolute calibration of the colour index and O₄ absorption derived from Multi AXis (MAX-)DOAS measurements and their application to a standardised cloud classification algorithm, Atmos. Meas. Tech., 9, 4803–4823, https://doi.org/10.5194/amt-9-4803-2016, 2016.

Zhao, X., Bognar, K., Fioletov, V., Pazmino, A., Goutail, F., Millán, L., Manney, G., Adams, C., and Strong, K.: Assessing the impact of clouds on ground-based UV–visible total column ozone measurements in the high Arctic, Atmos. Meas. Tech., 12, 2463–2483, https://doi.org/10.5194/amt-12-2463-2019, 2019.