Low-Cost Air Quality Sensors: One-Year Field Comparative Measurement of Different Gas Sensors and Particle Counters with Reference Monitors at Tušimice Observatory
Round 1
Reviewer 1 Report
The paper covers a very important issue of the reliability of environmental assessment based on low cost sensors.
The manuscript is well structures and well written.
I would ask Authors to address the following small issues:
- What was the actual percentage of the valid sensor measurement data, for the individual sensors during subsequent months of the year?
- L. 286. Do not you think that small R=-0.16 is considered statisticaly significant due to the lagre number of data, which makes that significance a bit dubious?
- Table 5. PM2.5 estimate based on PMS7003 is quite different from those based on Fidas200 and RM. But, In the text of the manyscript a coheremce was indicated. Please revise.
- L. 357-8. Does Fig S1 present the sensor drift.
- Was drift correction appllied before preparing descriptive statistics of sensors responses?
- Correlation analysis - was it based on 1 h average values in all cases?
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 2 Report
As a researcher who regularly performs small-payload high-altitude balloon flights -- not primarily for air quality measurements specifically, but for astronomical calibration of atmospheric optical density -- I have very often wondered why there does not (yet) exist a small, lightweight, property calibrated, and robust low-cost sensor (or for that matter even a reasonably small set of lightweight and accurate sensors) that can measure a large spectrum of atmospheric properties, including a decent set of trace gases, and particulates of various sizes, in addition to just the usual humidity and barometric pressure. Such a miniature, broad-spectrum, and well-calibrated sensor is most definitely not outside the range of present technological capability, yet they very sadly are not industrially available yet. Such sensors could, of course, be placed on standard lightweight radiosondes, small drone aircraft, as well as on ground-based weather stations and on commercial aircraft, etc etc etc. (A simple multi-trace-gas sensor operating at 5V or 3.3V that could provide data to any standard single-board computer [e.g., an Arduino or Raspberry Pi] via an I2C or SPI interface is what I personally would like to see.) Thus, I was very happy to receive this well-written and useful article that will provide much-needed feedback to sensor manufacturers toward the continued development of lightweight, broad-spectrum, and accurate sensors. The article appears to me to be very careful, thorough, and well-written, and I congratulate the authors for doing the work and writing the paper.
The only thing that I think could potentially be somewhat useful (and simple) to add, and perhaps of interest to the manufacturers working to improve their sensors, is a link within the supplementary material to files or a directory containing the raw data from the sensors that was used in making each of the plots and tables. The plots and tables in the article and supplementary material are thorough, but I think the raw sensor data itself could also potentially be of use to the sensor manufacturers.
Congratulations again to the authors.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 3 Report
This manuscript reports on a year-long field measurement, during which several low-cost gas and PM sensors were compared to reference methods. I agree with the authors that low-cost sensors could play a significant role in air-quality monitoring, provided that their performance is carefully characterized and their artefacts understood and corrected for. I believe that this study provides data that could be useful to the community and therefore I support the publication of the manuscript in the journal Atmosphere after minor revisions.
Comments
Page 2, line 56: The authors state that “there are no regulatory legislations or standards for quality control”. This is true, but there are efforts within the CEN the European Committee for Standardization (CEN) towards this direction. See for instance:
https://standards.cen.eu/dyn/www/f?p=CENWEB:7:0::::FSP_ORG_ID:2012773&cs=1FD71819F25D74834BB38751B78ACE16D
Page 5, Lines 181-182 and 199-200: The authors state that the measured data were “cleaned of any outages, negative values and significant outliers”. It would be useful, in my opinion, to report in sections 3.1 and 3.2 the % fraction of the PM data removed from the analysis for each sensor. (Indeed, in Section 3.2 the % fraction of the PM data characterised as outliers is reported for the OPC—N2 sensor. It would be nice if such information were provided for all other sensors).
Page 6, Section 3.1, Lines 213-214: The authors report a drift of 60 ppb and 10 ppm in the case of SO2 and CO sensors. Could you please also calculate and report the drift in %?
Table 3 and Table 5: I would suggest to add a footnote explaining what SD stands for (I assume that it is “standard deviation”).
Table 4 and Table 6: Please explain what “Presence” is.
Figure 3: Please explain what the black trace represents. I assume this is fit to the experimental data. Please provide more detailed information.
I would also suggest to change the axis labels to: O3 concentration RM (ppb) and O3 concentration Cairclip (ppb).
Page 12, Lines 384-385: The authors found that the max. lifetime of the Cairpol sensors is about 11 months. I was wondering how this compares with the manufacturer’s claims.
Sections 3.1 and 4.1: I find it rather confusing that the same gas sensors are sometimes referred to as “Cairclip” and sometimes as “Cairpol”. Even within the same table, e.g. Table 3, the sensors are referred to as Cairpol in the table caption, but as Cairclip in the table contents. I would suggest to be more consistent. Also, in Tables 3 and 4, there is a typo: the NO2 sensor is called “Carclip”.
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Reviewer 4 Report
The study describes the evaluation of a comparison of the observations recorded during a long monitoring campaign by both low-cost sensors of gases and particulate and the corresponding reference monitors. In my opinion the study is interesting because explores the different faults that can occur during a long-term monitoring campaign using low-cost sensors.
Nevertheless my main remark is about the graphs: as the authors would like to show if the low-cost sensors operate in the same way as the reference monitors, I strongly suggest to produce all the graphs presented in Fig.3,4,5,6 (and also in the supplementary material) with the same scale for both the x-and y-axis. I think that the visual inspection of the graphs by the reader can be improved in this way, as, for example, the graphs present in the paper by Bulot et al. [ref.15].
Otherwise, the authors can explain why they chose different scales for the the x-and y- axis.
Other observations follow:
Introduction
I think that the introduction is clear and contains adequate references. The scope of the study is clearly stated.
lines 55-56: please check English
lines 69-70: please check English
Experimental Methodology
lines 177-179: Is there a document to be cited in the references to confirm this statement?
line 185: please give a motivation about the choice of 10 minutes (conformity with RMs, other?)
Results
lines 232-234 (and fig. 3): it is known that ozone higher concentrations are usually found in the summer period thus is not so surprising to see this season-dependent different behaviour of the correlations considering also that the limit of detection of Cairclip is 20 ppb as stated in Tab.1. Moreover, observing the graphs presented in Fig. 3 it seems that at the operative conditions during the campaign the LOD of Cairclip is nearer to 40ppb than 20 ppb (see for example month 1 and 12). I suggest to the authors to check and comment a possible better correlation comparing values only if RM recorded >30-40 ppb because it seems that the curve assumes a possible linear relationship above those values, considering graphs about months from 4 to 9.
About PM correlations: it is known that RM with beta-attenuation and RM with gravimetric analysis are considered equivalent. On the other end the conversion from number of counts recorded by OPC to ug/m3 is not a trivial task because an estimation of PM density is needed, nevertheless many manufacturer include in the OPC software algorithms of conversion. See as an example Colombi et al. (Atmospheric Environment 70 (2013) 166-178, doi: 10.1016/j.atmosenv.2013.01.035). Have the authors any comments about this with respect to the results of their study?
Author Response
Please see the attachment.
Author Response File: Author Response.docx
Round 2
Reviewer 4 Report
I am satisfied by the authors replies to my comments, I think that the manuscript quality has been improved, thus I suggest that the manuscript is suitable for publication in the present revised form.