the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 17 Oct 2023
Development of compact continuous measurement system for atmospheric carbonyl sulfide concentration
Sebastian O. Danielache
Shigeyuki Ishidoya
Takahisa Maeda
Shohei Murayama
Abstract. Carbonyl sulfide (COS), the most abundant sulfur-containing gas in the atmosphere, is a source of stratospheric sulfate aerosol (SSA) and is used as a tracer for gross primary production (GPP). However, tropospheric COS sources and sinks entail great uncertainty due to the limited number COS observation sites. Thus, field measurements of COS concentrations worldwide are necessary to estimate the contribution of SSA and the global scale of GPP. Recently, MIRA Pico, a portable continuous COS concentration analyzer using mid-infrared absorption, has been released. MIRA Pico has a lower cost and is smaller than conventional laser COS analyzers. We modified and tested the MIRA Pico for atmospheric COS concentration measurements. The modified MIRA Pico exhibited ± 7.9 picomol (pmol) mol−1 (1σ) for a 15-min average, and calibration gas consumption was as low as no more than 3 L d−1. We also used the modified MIRA Pico for observations at Tsukuba, Japan. The observed COS concentrations ranged from 425 to 604 pmol mol−1, averaging a standard deviation (1σ) of (505 ± 33) pmol mol−1. The observed values agree with previous observations and exhibit clear diurnal variations. Furthermore, we installed the modified MIRA Pico in a passenger car to observe the COS concentration distribution in Tsukuba City.
- Preprint
(2522 KB) - Metadata XML
- BibTeX
- EndNote
Kazuki Kamezaki et al.
Status: final response (author comments only)
-
RC1: 'Comment on amt-2023-209', Marc von Hobe, 09 Nov 2023
Overall evaluation and general comments:
The preprint by Kazuki Kamezaki on a “compact continuous measurement system for atmospheric carbonyl sulfide” describes and evaluates modifications to improve a commercially available OCS instrument. In that respect, the term “Development of…” in the title is clearly an overstatement. While this could be amended by changing the title, I’m afraid that the paper in its present form contains too little substance and too many errors and ambiguities to be publishable in AMT. The necessary modifications go beyond normal revisions, which is why I recommend not to accept this paper for publication. Below, I explain my main concerns. Then I list some specific and technical issues to underline my criticism and to provide recommendations should the authors chose to rewrite the article for resubmission at a later stage.
The methods section (Section 2) falls short of providing anything close to a comprehensive instrument description. Neither is the overall measurement and data analysis concept of the original MIRO analyzer fully described, nor are shortcomings or problems that motivated the adaptations made to the analyzer in this work properly characterized. In a methodological study such as this one, I expect, for example, the following details to be given and explained: What are the dimensions of the optical cell and what is the absorption path length? What light source and detector are being used? At what wavelength is OCS being measured? Does the analyzer measure an absorption spectrum that is analyzed by spectral fitting, or does it measure absorption only at one or more distinct wavelengths? Which water absorption band is used as a reference peak, and how does this referencing work? How is the temperature of the light source stabilized, and how is wavelength and baseline stability ensured? What are the main sources of noise? Without such methodological details, it is impossible to put the applied modifications into context and evaluate their benefits as well as the overall performance measures.
With respect to performance measures (given in Section 3) some of the numbers are neither well rationalized nor convincing. The terms accuracy, precision, and uncertainty are not always used in a correct and consistent manner, and it is not at all clear how the results of the various measurement series performed on know standards are used to validate or correct OCS concentrations measured in the field.
The description of the field measurements in Section 4 is not very meaningful in my opinion. The explanations and interpretations are largely rudimentary and speculative. They go too far for a purely technical paper, but not far enough in order to be of scientific value. Given the time period reported and the region covered, the results are only of (rather limited) local interest, and the case that more such measurements could be a game changer in understanding the OCS budget and cycling is not convincing.
Specific comments:
Line 27 – 30: In my opinion, these two sentences are misleading. First, there are no “local contributions to SSA production”, which happens in the stratosphere and is not directly connected to the near-surface OCS cycling. What you mean is contributions to the budget, which in turn plays a role for how much OCS reaches the stratosphere, where SSA is produced. More importantly, the statement that “tropospheric OCS sources and sinks entail great uncertainty due to the limited number COS observation sites” is far too simplistic. Arguably, a few more sites with OCS observations in well chosen locations could help to better quantify the regional distribution of certain sources and sinks. But overall, a great deal of understanding tropospheric OCS cycling has been achieved with data from available networks and satellites, and I expect a bit more in terms of strategy to address remaining uncertainties than simply calling for more OCS observations.
Lines 39 + 47: The statement “However, these devices are large, expensive, and costly to maintain.” is too unspecific as these adjectives are somewhat relative. Also, a few lines below, you define a “good precision” out of the blue. I would like to see more compelling arguments what the precision needs to be to tackle relevant science questions, and what the exact problems are with the size and costs of existing analyzers. In other words, where are the real problems that need to be solved? Later, it is stated that the MIRO is “less than half the price” of other analyzers and “small”. Again, please be more specific here (at least make some reference to the details given in Section 2) and clearly state why the differences matter.
Line 58 – 59: If > 5000 ppm of water vapour are needed for the OCS measurement to work, then the instrument will be useless in cold or high-altitude environments (unless you go through the not simple efforts to moisten the sampled air). This should be clearly stated.
Line 66 – 69: From the text and Figure 1, I find it difficult to understand how the temperature stabilization was really done. Can you give details on the “commercial refrigerator”? And was the entire MIRA instrument put into the refrigerator or just the optical cell? I also don’t understand why you would set the refrigerator to 15 °C and the Peltier cooler to 29 °C? What is the target temperature inside the optical cell, and is temperature monitored inside the cell or just inside the refrigerator or at the Peltier cooler? And what is the "cushioning material" and what is its purpose? If it is thermal insulation, then “cushioning” is the wrong word.
Section 2.2: There are several things that I don't understand in this section and that need further explanation and rationale: What is the purpose of the ECU and cooling the sample to 2 C? How do you know/ensure that the fraction of OCS removed by the activated charcoal is constant? With the Nafion dryer, how do you manage to keep the > 5000 ppm water to make the MIRO work (cf. Section 2.1)?
Line 92: How do you “humidify” the dry standard gases by passing them through and ECU set to 2 °C? If there is any amount of water present, cooling the gas would either remove water due to condensation or increase the relative humidity. If no water is removed, the absolute amount or mole fraction should remain constant.
Section 2.4: This Section completely lacks the necessary detail. For the little information given here, I can only guess that the MIRO gives out some offset OCS concentration even for samples that do not contain any OCS, and that you assume the reference gas to be OCS free so that it can be used to quantify this offset. But I’m missing proof that this is really the case (see next comment).
Section 3.1: I have several issues with this section: (i) Why are Alan deviations of the original and modified systems compared for different time constants (5s/160s vs. 40s/180s)? (ii) If I understand it correctly, the comparison was not made for the same sampling procedure, i. e. there were no reference gas injections with the original instrument. The rationale behind this and why you don't think that the results are affected by this should be explained. (iii) When using room air with an activated charcoal filter that does not remove all the OCS, how can you rule out OCS variability in your reference gas? In other words: how can a gas with a low but unknown OCS concentration serve as a reference? (iv) When looking at the Alan deviation plot in Figure 2, it becomes evident that the original instrument performs significantly better in terms of precision at time constants below 10 seconds. This needs to be discussed! When looking at the Alan deviation for longer time constants and considering your modifications, the drifts in the original system appear to arise from temperature instability. This should be discussed. If the red curve is correct, one take home message of your work is that the original MIRO has severe problems with temperature stabilization that render it useless for long term measurements!
Section 3.2: In this Section, you demonstrate good linearity (Figure 5) and the absence of significant long-term drifts (Figure 4) but that doesn’t necessarily translate to good accuracy. It is evident from Figure 4 that, on average, the MIRO tends to underestimate the concentrations of all three standards, which is also reflected by the slope of the calibration curve being 0.90 rather than 1 and the intercept being negative rather zero. I'm missing a discussion why this is the case, and a description if and how the calibration curve is used to correct observed concentrations.
Line 137: “This shows that MIRA Pico programmatically corrects for water content.” I find it strange that you need to demonstrate this. Whether the instrument makes such an internal correction or not should be information available from the manufacturer. Tests should only be necessary to evaluate it this works or not. Ideally, if there is an internal water vapor correction, information should be given on how exactly this is done!
Lines 140 – 143: What I see in Figure 6b (and also in Figures 3 and 4) makes it hard to believe the numbers stated in the text. The test period was 6 weeks, so the individual measurement points are hours to days apart, and offsets from the mean or target values for individual points appear to be more on the order of 10 – 30 ppt. I would really like to see the math behind deriving the stated overall uncertainties.
Line 151 – 152: Why do you state "overall uncertainty" (which I assume to combine accuracy and precision) and "repeatability" (with which I assume you mean precision) for different time constants? It is not clear to me what you really want to say.
Line 154: It seems odd that ±0.5 °C should have such a strong effect on precision. What time scales are we talking about here? And it is laser/detector stability, or is it a temperature effect on the gas concentrations? It could be useful in this context to show the measured temperatures, and possibly try to correlate them with any deviations between observed OCS and the known values from the standards.
Line 157 – 164: Are the calibration gas flows used in the cited studies really necessary, or could similar results be achieved with less gas consumption? After all, it is at least possible that the other groups didn’t make the strongest efforts to minimize calibration gas flows.
Section 4.1: The interpretations of the observations sound rather speculative. I'm not saying that the explanations are necessarily incorrect, but it is just not a real interpretation, as is evident from the last sentence (line 196 – 198). I suggest to either do it right (which should not be too difficult for a 10 day period, and you obviously have local weather data at hand) or leave it out of the paper.
Line 211 – 215: I wonder what the point is in listing these exact wind speed observations. I suggest to only mention and discuss wind speed in the context of actually trying to explain observed OCS (as attempted in the following paragraph).
Line 227 – 229: What does the observed OCS distribution have to do with measurement accuracy?
Line 230 – 236: There are surely applications where a small, mobile OCS analyzer could be very useful, so efforts to build such a device that produces reliable data are clearly warranted. However, to understand local source and sink processes and overall cycling of OCS, it would be much more useful to measure OCS at the most interesting sites for several days at a time. For a chemically stable gas like OCS, transport and boundary layer dynamics play a key role for local variability, and these are impossible to analyze when taking the instrument from one place to another in a fast moving car.
Line 239 – 240: The Alan deviation only decreased for time periods larger than 10 seconds.
Line 249 – 251: The "expected" concentrations do not belong into the conclusions. Summarize important actual observations if there is anything substantial to report.
Technical issues:
I always find “pmol mol-1” awkward for mole fractions. Please consider using ppt instead, which is widely used and understood.
Line 15 and line 247: „…averaging a standard deviation (1-σ) of (505 ± 33)…” seems as if 505 is meant to be the standard deviation. The average is 505 and the 1-σ standard deviation is 33, and this should be reflected in the wording.
Line 24: it should be “ozone depletion” and not “ozone depression”
Line 24 – 26: the grammar in this sentence appears to be incorrect.
Line 37: The LGR analyzers use off-axis integrated cavity output spectroscopy, not cavity ringdown spectroscopy (there are important differences between the two).
Line 62: What is meant by “standard deviation of 10-min accuracy”? Accuracy reflects systematic errors or biases and should not depend on any time constant. And what should the standard deviation of the accuracy be relevant for?
Line 204: Do you mean 1000 Wh for the batteries?
Line 212: There is no Supporting info with this preprint!
Line 254: The last sentence is obviously incomplete.
Figure 3: one zero needs to be removed from the numbers on the y-axis of panel a.
Figure 9: The latitude and longitude panels are not helpful here, as the covered ranges are too small to have any significant effect. To relate the time series to the track on the map, marking a few way points or landmarks by vertical lines in the time series and markers on the map seems the more intuitive solution.
Citation: https://doi.org/10.5194/amt-2023-209-RC1 -
RC2: 'Comment on amt-2023-209', Anonymous Referee #2, 11 Nov 2023
General assessment:
Kamezaki et al present and test a new mid-cost analyzer (MIRA pico) for carbonyl sulfide (COS) concentration measurements. They also test a modified measurement setup in order to reduce the drift of the analyzer and report COS concentration measurements from Tsukuba, Japan.
COS measurements have gained more attention during recent years, especially because of the link between vegetation COS exchange and stomatal conductance and/or photosynthesis. However, the global COS budget is still not closed, partly due to the lack of a proper COS measurement network. As the mostly used gas analyzers (from Aerodyne and Los Gatos) are high cost, they are not that widely implemented. In the light of missing COS data, a mid-cost analyzer is very welcome to improve COS data availability. Given that the accuracy of the presented instrument is not very good compared to the Aerodyne and Los Gatos analyzers, I don’t really see it as an option for atmospheric concentration measurements, that typically require very high precision. Instead, I would more likely see this type of portable analyzer convenient for e.g. chamber flux measurements. This study tests the long-term stability and reference gas consumption of the MIRA pico analyzer. However, the purpose of the use of reference gas is unclear, since the authors mention using room air as reference, that has an unknown amount of COS. The manuscript by Kamezaki et al still lacks many crucial information related to e.g. the measurement setup and the operation of the instrument they present, leaving the reader with many unknowns. The manuscript thus needs very substantial revisions and should either go through (very) major revisions, or rather rejected and resubmitted after modifications. I have listed the specific shortcomings below, followed by the detailed comments.
- More information is definitely needed on the instrument itself: what is the size of the sample cell? How is COS concentration measured, at what wavelength and why does it need water vapor? Why is activated charcoal needed? Why a nafion dryer is installed if the measurement itself needs a certain amount of water vapor? How to know if there is enough water vapor for a reliable measurement? I also don’t understand why there is an ECU to first humidify the sample air and after the ECU there is a nafion dryer to dry it…? What exactly is the refrigerator and how the sample cell can be moved there? What is the overall size of the modified system, is it still portable? How can indoor air be used as reference gas as it has and unknown COS concentration? The authors need to be more specific on these details.
- Some explanations are needed, e.g. Why is sample and reference gases switched every 30s? Is it really necessary that frequent, is it sustainable?
- Temperature stability is mentioned, but results are not shown. I suggest to add more results related to temperature stability and effects on COS concentration, at least in a supplement.
- The field measurements are not described at all in the methods section, so it is very difficult to assess their relevance.
- From the Allan variance plot it is clear that the low-frequency drift decreased after the modification to the measurement system. However, the high-frequency noise was increased. This is not discussed at all in the manuscript or the reasons why this increase happens. It is also unclear why the optimum integration time (for highest accuracy measurements) then changes from 10s to 40s?
- Trajectory analysis would be needed to know where air parcels actually came from to better analyze and give relevance to Fig. 8. The whole field measurement section is lacking supporting information and either needs to be expanded or left out.
Detailed comments:
Figure 1: Where is the outlet from the analyzer? Did you take the measurement cell out of the analyzer and put it in the refrigerator..? You mention in the text the size of the analyzer but what is the size of the modified setup? A picture of the setup would also be nice, e.g. in supplement
Figure 2: Gridlines would help the reader. From low frequency variation you can determine if the drift is linear or non-linear, please do that either in the text or also show lines of linear and non-linear drift in the plot as in e.g. Gerdel et al. 2018. In the caption: “Allan deviation plots with original…”
Figure 3: Why is there an (more or less) empty area in the middle of the scattered measurements…? As if the analyzer could not detect certain concentrations, only the scatter. Panel a COS concentrations seem to be 10 times too high. What exactly is the difference of plots a and b? Averaging? “The plot was almost every second” what does this mean? Do you mean the frequency of the measurements was 1 Hz?
Figure 4: Why is there data missing May 1 to May 10th and then again for a few days..? What is the time scale of these measurements? Please add the concentrations of the standards e.g. as lines to the plot or in the figure caption. Why are all Standards measured as 50-60 pmol mol-1 lower than what they should be?
Figure 5: Please add panels a and b, and refer to them in the caption. Why are there dots inside the circles in the lower (b) panel? Why are error bars omitted?
Figure 6: There is a big gap from May 1st to May 10th and then some days again, and after this gap there is a big step change especially for Standard A concentrations. What happens here? Please explain in the text. Why is this step change not visible in panel b? Did COS concentration also have this step change..?
Figure 7: I suggest to move this fig to a supplement
Figure 8: Meteorological variables would be very beneficial in interpreting this figure. Please add at least wind speed and direction as well as air temperature and relative humidity time series plots to this figure.
Figure 9: Averaging time 15 min is mentioned twice, please check. It would be informative if you plot all original datapoints (maybe in lighter color) and then the averaged values on top in panel a.
Figure 10: Is “Tsukuba site” the same as “swamp”..? Please make clear and be consistent. The scale on the lower right corner should be more visible. You could mark urban areas e.g. with rectangles/circles in the maps.
Abstract: Mention the manufacturer (Aeris Technologies) of MIRA Pico somewhere
L25: “carbon dioxide (CO2)”, as this is the first time CO2 is mentioned in the text
L28: “…limited number of COS observation sites.”
L34: Aerodyne quantum cascade laser spectrometer (QCLS) (Aerodyne Research Inc., Billerica, USA)
L36: Kooijmans, not Kooijimans
L37: ABB-LGR off-axis integrated cavity output spectroscopy (OA-ICOS)
L47: “less than half that of a conventional COS analyzer”: How much is it with the modifications you made?
L55: “carbon dioxide (CO2)” -> “CO2” as you should already introduce CO2 in L25
L55: “water vapor concentrations”
L59-L61: These two sentences are more like introduction than methods; suggesting to move to Introduction.
L62: “standard deviation (1 σ) of ± 50 pmol mol-1 on 10 min time scale”
L74: What was the material of filter and inlet tubing?
L80: Is this shown somewhere, that there is no difference? Why are pump and ECU then used if there is no change? This needs some rephrasing.
L83: “Activated charcoal can remove a part of COS.” This sounds very dangerous, why would you want to remove some of the target gas..?.
L90: You mention when the standards were filled, but not when were the lab measurements and long-term stability tests done? From Fig 4 I see in spring 2023, but mention it also in the text in Methods section
L95: If sample air and reference gas are switched every 30s and data are collected only during 10s, that means only 20s of actual data remain every minute..?
L104: “..many studies have reported that..” please provide references
L115-116: “some level of fluctuation”, please quantify how much (e.g. 1 min std). What time frame is the std for standard C presented?
L119: “countenious” -> “continuous”
L135-139: It was mentioned previously that a water vapor concentration of at least 5000 µmol mol-1 is needed for COS measurements, but you report a water content of 4000 µmol mol-1 for Tsukuba. Are the COS measurements then unreliable? How is the water vapor concentration measured after humidifying and drying the air?
L148: “Koiijimans” -> “Kooijmans”
L160-164: The amount of reference/calibration gas used depends on the user and target of the measurement (e.g., frequent calibrations are not as necessary for flux measurements as they are for accurate atmospheric concentration measurements), not only on the analyzer used. Kooijmans et al. (2016) measured a reference gas every 30min for 3min, not for 10min, so this estimate of their reference gas use is quite misleading.
Sect. 4: I suggest to rethink the organization of the sections, since this section is still very much about results and discussion (sect. 3). I suggest to change the numbering of this section from 4 to 3.5 and the subsections as 3.5.1 and 3.5.2.
L186-188: Sentence beginning with “They decreased..” and the following sentence: Please rephrase these sentences as they are not very clear. One suggestion would be “COS concentrations increased after sunrise until approximately 16:00, after which they decreased.” I would also suggest a plot with average/median diurnal variation. It is quite difficult to determine from Fig. 8; e.g., it seems on 19th April the decrease would happen at 18:00 while on 20th April it happens only after midnight .
L192-194: The decrease of the atmospheric concentration is especially because of the atmospheric mixing conditions, and since you observe a decrease during nighttime it means there is a nighttime sink in the ecosystem (e.g. soil bacteria as you suggest). Concentrations increase again after sunrise as the atmospheric boundary layer increases and mixing layer develops, mixing higher concentrations from above with the air close to surface.
L199: I would also suggest flux measurements (either by chambers or eddy covariance) to determine the sinks and/or sources.
L207: “Almost the same COS concentration was observed”: please elaborate this, especially close to swamp the concentration is considerably different
L210: Where are the urban areas located? Not really visible from the map.
L214: Since the wind was from south-southwest, could it be there is COS signal from the industrial area?
L214: Why 14:00-14:20 is selected as an interesting timeframe?
L225: Photoproduction from wetlands but also consumption by photosynthesis has been reported in previous studies (see synthesis study by Whelan et al., 2018)
L235: Why only areas with high or low COS concentration are interesting? How to even know that without measuring?
L239: Allen -> Allan
L240: Allan variance was decreased only at low frequency, but at high frequency it actually increased!
L244: “smaller cell”, I think it was not mentioned how big is the cell? Please mention it in the Methods section
L250: “was expected” but was it shown/reported?
L254: Again, I don’t know why COS concentration measurements would only be interesting in areas with high or low concentration
L254: The last sentence is incomplete.
References:
Gerdel, K., Spielmann, F. M., Hammerle, A., & Wohlfahrt, G. (2017). Eddy covariance carbonyl sulfide flux measurements with a quantum cascade laser absorption spectrometer. Atmospheric measurement techniques, 10(9), 3525-3537.
Kooijmans, L. M., Uitslag, N. A., Zahniser, M. S., Nelson, D. D., Montzka, S. A., & Chen, H. (2016). Continuous and high-precision atmospheric concentration measurements of COS, CO2, CO and H2O using a quantum cascade laser spectrometer (QCLS). Atmospheric Measurement Techniques, 9(11), 5293-5314.
Whelan, M. E., Lennartz, S. T., Gimeno, T. E., Wehr, R., Wohlfahrt, G., Wang, Y., ... & Campbell, J. E. (2018). Reviews and syntheses: Carbonyl sulfide as a multi-scale tracer for carbon and water cycles. Biogeosciences, 15(12), 3625-3657.
Citation: https://doi.org/10.5194/amt-2023-209-RC2
Kazuki Kamezaki et al.
Kazuki Kamezaki et al.
Viewed
| HTML | XML | Total | BibTeX | EndNote | |
|---|---|---|---|---|---|
| 172 | 53 | 5 | 230 | 3 | 4 |
- HTML: 172
- PDF: 53
- XML: 5
- Total: 230
- BibTeX: 3
- EndNote: 4
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1