the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
| 14 Jul 2026
Portable system for characterizing greenhouse gas analyzers for sensitivity to water vapor
Abstract. Water vapor introduces dilution, spectral interference and pressure-broadening effects in greenhouse gas (GHG) measurements using infrared (IR) techniques. Accurate water vapor corrections are therefore required for measurements of ambient air samples that are undried or incompletely dried as is the case for commonly used permeation water exchangers. In this study, we developed a lightweight, low-cost, and portable system for performing water vapor corrections of GHG analyzers in both laboratory and field environments. The system uses a permeable membrane-based moisture exchanger (Permapure, Inc., BE series) to humidify gas from a compressed cylinder with known analyte mole fractions. This humidified stream is mixed with a parallel dry stream from the same cylinder to produce air with controlled water vapor levels. Water-vapor (H2O) concentrations are held constant for a defined interval and then stepped across a range of user-defined levels. Gas flow rates are regulated using mass flow controllers, and a Raspberry Pi-based Linux system automates the sequence of water vapor steps and records measurement data for subsequent analysis. Compared to previously available dew point generators, the system provides substantially improved water vapor stability (approximately 3 ppm H2O). The system was used to characterize water vapor corrections for Aeris MIRA Ultra and Picarro cavity ring-down spectroscopy (CRDS) analyzers. The two tested Aeris MIRA Ultra analyzers exhibited quadratic relationships for methane (CH4), with distinct behaviors observed between low- and high-H2O regimes, with errors of up to 30 ppb CH4 at 3 % H2O if uncorrected. However, these two tested Aeris MIRA Ultra analyzers did not show clear dependence on water vapor for ethane (C2H6). Across the investigated humidity range, all three tested Picarro analyzers showed a nearly linear dependence for carbon dioxide (CO2), and both analyzers tested for CH4 exhibited a similar near-linear response. For carbon monoxide (CO), however, two of the three tested analyzers displayed a quadratic humidity dependence, and one did not show any dependence on water vapor. For the Picarro analyzers, the errors if uncorrected for water vapor response were as large as 0.25 ppm CO2, 2.0 ppb CH4, and 12 ppb CO, all at 1.7 % H2O. After applying the derived analyzer-specific water vapor correction functions to all datasets, the corrected CH4, CO2, and CO measurements from Aeris MIRA Ultra and Picarro analyzers were well within the WMO/GAW inter-laboratory compatibility goals. These results demonstrate that the system provides a practical and reliable approach for conducting water vapor corrections for GHG analyzers in laboratory and field applications.
Competing interests: The authors declare that a provisional patent application related to the P-WAVES system described in this manuscript has been filed by the Pennsylvania State University.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.- Preprint
(2511 KB) - Metadata XML
-
Supplement
(660 KB) - BibTeX
- EndNote
Status: final response (author comments only)
-
CC1: 'Comment on egusphere-2026-3732', Andrew Whitehill, 31 Jul 2026
-
AC3: 'Reply on CC1', Yunsong Liu, 17 Oct 2026
We would like to thank the referee for the review and comments to improve the quality of this work. Following these comments, we have made major modifications and point-by-point responses. Please find our response below (in blue).
The manuscript “Portable system for characterizing greenhouse gas analyzers for sensitivity to water vapor” by Liu et al. claims to describe a “lightweight, low-cost, and portable system” for performing water vapor corrections of greenhouse gas analyzers.
The basic premise of their system is forming a “wet” stream of calibration gas using a Nafion-based humidity exchanger, then combining it with a “dry” stream at set ratios using mass flow controllers and valves. This is a description of common laboratory practice that has been used for decades in air pollution / air quality and VOC laboratories and in instrument development and testing, in commercially available instruments (e.g. 2B Technologies DewLine), and presumably also used already in the greenhouse gas community.
We thank the reviewer for this comment highlighting the need to clarify our contribution and we changed Lines 70-77 in the introduction to read “Using Nafions to condition air samples (either stabilizing the humidity, drying, or equilibrating to a reference humidity) is certainly not new and has been used in many fields. For example, the 2B Technologies DewLine (https://2btech.io/dewline/) uses a Nafion tube to equilibrate humidity of the sample and reference gas streams. In our instruments deployed in the field (e.g. Richardson et al., 2017; Liu et al., 2025), we also use a Nafion to equilibrate the humidity (in our case to low levels, around 1000 ppm H2O). Because of the extreme compatibility goals for greenhouse gases, this procedure of drying the air sample still leaves some humidity effects on the final results. In this paper, we introduce a new portable and automated system that enables water-vapor calibrations to be performed directly in the field. We use this system to develop calibration curves to adjust for those relatively small effects in order to improve overall compatibility.”
The concept of mixing humidified (using a Nafion tube submersed in water) and dry split flows using mass flow controllers has also been documented in the greenhouse gas community literature (see Lines 61-63 in the introduction of the paper). Another example is Dembowski et al. (2021), which documented the use of this concept to investigate the role of relative humidity in determining which aluminum mineral forms under various conditions. They use feedback from a humidity sensor to ascertain the humidity levels remain constant – this would be a potentially valuable improvement for future iterations of our system. We added this point to the discussion (Lines 306-310).
References
Dembowski, M., Loring, J. S., Bowden, M. E., Reynolds, J. G., Graham, T. R., Rosso, K. M., and Pearce, C. I.: The controlling role of atmosphere in dawsonite versus gibbsite precipitation from tetrahedral aluminate species, Dalton Trans., 50, 13438–13446, https://doi.org/10.1039/d1dt02081a, 2021.
The claim to novelty, then, comes from the weight, cost, portability, and apparent stability of their system. They provide dimensions (30.5 cm x 30.5 cm x 20.3 cm, 5 kg) but it is not clear whether that includes all parts of the system (including the water and a battery) or not. They claim “low power requirements” but do not specifically test or report the power draw or make any claims for how long a typical battery-operated system could run in the field. They also provide no head-to-head comparisons of their system to existing, commercially available systems to provide an equivalent comparison.
We agree that using mass flow controllers and a permeation-based water exchanger is not novel and have clarified Lines 70-71 in the manuscript. As noted by the reviewer, the novelty of the system is the weight, cost, portability, and stability. We have also clarified specifics as noted below.
We have revised the manuscript to clarify what is included in the reported 30.5 × 30.5 × 20.3 cm dimensions and 5 kg system mass, and to distinguish the present configuration from an optional battery-powered configuration (Line 100).
We thank the reviewer for this comment and agree that the original manuscript did not provide sufficient quantitative information regarding power consumption and battery operation. We have now measured the steady-state power consumption of P-WAVES during normal operation and found it to be approximately 10 W. Thus, a 24-h field deployment would require approximately 240 Wh of delivered energy. As a representative example, a commercially available 12-V, 30-Ah LiFePO4 battery provides sufficient capacity for more than 24 h of operation while allowing reasonable reserve for conversion losses. Such a battery is approximately 16.5 × 16.3 × 8.6 cm, weighs approximately 3.2 kg, and costs less than US$300. It could readily be accommodated within the enclosure currently used for P-WAVES. We added these important details to Lines 107-112 in the manuscript.
We also agree that the original discussion could be interpreted as implying a direct performance advantage over commercially available humidity-generation systems. A head-to-head experimental comparison was not part of this study, and we have revised the text accordingly. Our intended novelty claim is not that P-WAVES outperforms commercial humidity generators in absolute metrological performance in terms of accuracy, but rather that it provides a small, low-cost, low-power, automated system specifically designed for portable field characterization of greenhouse-gas analyzers across controlled stable H2O levels. Commercial systems may provide substantially greater environmental control, traceability, and absolute humidity accuracy, but generally address a different use case. We have therefore moderated the wording of the novelty claim and clarified that the principal advantages of P-WAVES are its portability, low cost, low power requirement, automation, and suitability for analyzer-specific field calibration (at Lines 75-77). RH sensors are accurate enough to hold H2O to ~500 ppm and chilled mirror sensor to within approximately 20-50 ppm H2O, so the stability of the P-WAVES system (2-3 ppm below 10000 ppm and 7-10 ppm up to saturation) is an improvement. While the P-WAVES system is very stable, its accuracy is untested, beyond using the Picarro and Aeris water vapor output which is itself not calibrated.
The reviewer is correct to point out that we do not have any head-to-head comparisons of our system to an existing humidity measurement (beyond that of the instruments to be calibrated). Using a sensor, such as a chilled mirror hygrometer inline would be a potentially valuable addition, particularly if the system is to be used for other applications. We added this point to the discussion (Lines 292-294). However, for our application, the key point is the value of water reported by the instrument, since, in practice, that is what we will be using in the field to correct for the remaining water vapor after using a Nafion dryer to remove most of the water. The actual value of the water and its calibration is not crucial (unless the calibration of the water reported by the instrument changes dramatically in the time between calibrations or during individual steps of the calibration).
The authors make a claim of “high stability” (3 ppmv H2O). However, in the manuscript they appear to only take the average of the last 60 seconds of a 3.5 minute step. It would be much more convincing if they reported a response time or equilibration time to humidity change and provided the long-term (e.g. 8 hour, 24 hour, etc.) stability of the system instead of 60 second transients. A 60-second transient does not show “high stability”. Depending upon the type of tubing used (polymer, stainless steel) and specific instruments used, the system response time to water vapor could be on the order of 60 seconds or longer. The apparent high stability could be a system-induced broadening kernel. The authors do not state what materials, tube diameters and lengths, etc. they used, nor do they provide any residence time distribution data to prove that their 60-second stability is actually the stability of their humidity generation and not a system artifact.
We agree that a 60-s interval characterizes short-term within-step stability rather than long-term stability, and have revised the terminology accordingly. For our application, within-step stability is more important but other applications may need long-term stability so we additionally performed a 16-h fixed-setpoint experiment to characterize the P-WAVES system for applications in which long-term stability is required. Slow changes in generated H2O were strongly correlated with the Aeris enclosure-inlet temperature (T2; R2=0.99 for 5-min averages after the first hour) and were consistent with the expected temperature dependence of water-vapor saturation (see figure below, now Figure SX in the supplement). During the thermally stable final 8 h (by coincidence) of the 16 h test, H2O drift was approximately -5 ppm h-1, while median 60-s variability was approximately 5 ppm (1σ). We added the plots in the supplement (Figure S6) and the text describing this test in the discussion (Lines 326-334). We have therefore revised the manuscript to distinguish short-term within-step stability from long-term absolute stability and to note the importance of temperature measurement or control for extended fixed-humidity operation.
Figure S6. (a) Water vapor as a function of time during a 16-h fixed setpoint test. Enclosure inlet temperature (T2), used as a proxy for water bath temperature, is shown on the right-hand y-axis. (b) Water vapor as a function of enclosure inlet temperature for the test shown in (a).
The system uses 1/8’’ OD stainless steel tubing (length is estimated to be about 50 cm of stainless tubing internal to the calibration unit) and we apologize for the oversight in not including this important detail. We do not in general consider longer time steps because our primary goal with the water calibrator system is to calibrate the Aeris sensors for water vapor and these instruments drift dramatically in time (Liu et al., 2025), particularly for ethane. Our aim is to keep the time steps short in order to isolate water vapor effects from instrument drift.
Additionally, we added a figure (below) to the supplement (Figure S1) to more carefully address the water vapor variability at various levels during the steps used to develop water calibration curves for the instruments. The variability increases with increasing water vapor for the completed tests. Across 29 humidity-step tests comprising 368 individual steps for two Aeris analyzers, the median final-60-s H2O variability was 2.8 ppm (1σ), with typical variability of ~2-3 ppm below 20,000 ppm H2O and ~7-10 ppm at the highest H2O levels.
Figure S1. Variability of generated H2O as a function of water vapor concentration and step duration. Each symbol represents the standard deviation of H2O during the final 60s of an individual concentration step. Different symbols indicate step durations of 3, 3.5, 4, and 5 min. The black line represents the binned median across all 3-5 min tests.
Across the 3-5 min humidity-step tests, final-60-s H2O variability showed no clear systematic dependence on step duration.
Based on this reviewer’s concern regarding the ability to see details within the steps (also noted by reviewer RC2), we added insets to Fig.2 in order to better visualize the stability of the steps. The stability is quantified by the reported standard deviation.
The authors claim “stable and repeatable humidity control…” but provide no statistics or demonstration of long-term stability and either short-term or long-term repeatability. As far as I can tell, their experiments were purely done in a climate-controlled laboratory environment. This performance might not transfer to field environments. In particular, the stability of the humidity generation will be highly sensitive to the temperature of the system, so ambient temperature fluctuations will adversely impact the stability of the system. It is not clear that the authors have thoroughly evaluated and addressed these impacts.
Indeed the tests were all done in a climate-controlled laboratory and the temperature of the laboratory (more specifically the water bath) dictates the saturation water pressure (i.e., the maximum water vapor attainable). As also noted by the reviewer, changes in the water bath temperature throughout each step would affect the stability of the generated water vapor level. An improvement to the system could be to measure the water bath (or use a chilled mirror sensor to measure the water vapor) and potentially automatically adjust the mass flow controllers to compensate. But without these complications, the laboratory stability of the water vapor within the steps is 4.8 ± 2.9 ppm H2O, sufficient for our laboratory characterizations purposes. (add to discussion Lines 310-312). Vacuum-sealed enclosure should be used for field tests.
Additionally, we have now examined whether temperature variability could account for the observed short-term H2O variability during the calibration steps. We compared the standard deviation of the H2O during the calibration steps to the standard deviation of the enclosure inlet temperature (T2, see below). The correlation coefficient (R2) was 0.006 for the 313 windows from the 3-5 min tests. Thus, variation in enclosure inlet temperature explains essentially none of the observed short-term H2O variability during the final minute of humidity steps. The short-term H2O variability may be caused by Aeris H2O measurement noise, MFC flow-control variability, and/or analog setpoint-voltage stability. We added this figure to the supplement (Figure S4) and the text to Lines 314-318 in the discussion.
Figure S4. Standard deviation of the H2O during the calibration steps to the standard deviation of the enclosure inlet temperature (T2) of 21 tests (313 windows) for Aeris MIRA analyzers (serial number A778 and A798). The correction coefficient (R2) was 0.006 for these 3-5 min tests.
For our application, the key is the reproducibility of the calibration curves for water vapor. Figure 3 and 5-6 show CO2 and CH4 repeated tests for the Aeris and Picarro systems. Table 1 and 2 document the related statistics and its repeatability is discussed in Lines 319-325 in the manuscript.
The authors do not use any independent humidity or dewpoint sensors (such as a chilled mirror or capacitive sensor) to confirm the accuracy or stability of their humidity generation. Instead, they rely on instrument-reported water vapor lines. However, they provide no calibration or traceability for these measurements. As a humidification system or humidity generator it has no inherent humidity measurement capability.
As noted above, the actual value of the water vapor is not crucial for our particular application. The reviewer is correct that humidity measurement capability should be added in order to use this system more generally and we have noted this in the discussion (Lines 290-294).
As mentioned earlier, much of the system (split flows, Nafion humidification, mass flow controllers to adjust humidity, etc.) are already common practice in the field. If the authors want to make a novel claim about the superiority of their system they need to provide additional details about the other factors (mechanical designs, CAD files, Python code, parts lists, etc.) to make it reproducible and independently testable. If the authors can address some of these concerns and provide more convincing data about the value of their system I think it would be a useful contribution to the field, especially if the technology is made available (public domain or commercially).
Thank you for the suggestion. We have made the Python control software used to operate P-WAVES publicly available on GitHub (https://github.com/srichard13/P-WAVES). The repository includes the production controller used for the Aeris MIRA Ultra analyzers and a representative Picarro controller (CFKADS2025). Separate Picarro versions were used for individual instruments because of instrument-specific communication/configuration differences. The archived source files are provided unchanged from the versions used for the experiments, together with documentation of the hardware assumptions and logged variables.
I do not address the Aeris and Picarro testing here, as those aspects are already well-covered in the scientific literature and well known to be dependent upon instrument, firmware / software version, etc. and not necessarily transferrable (although the calibration process and applied correction models could be).
We believe our system to be a significant improvement on the systems currently documented in the greenhouse gas literature (Lines 75-80 in the introduction) primarily because of the size, stability, and automation.
-
AC3: 'Reply on CC1', Yunsong Liu, 17 Oct 2026
-
RC1: 'Comment on egusphere-2026-3732', Anonymous Referee #1, 07 Aug 2026
Review: Portable system for characterizing greenhouse gas analyzers for sensitivity to water vapor.
General Comments.
- There is a clear need for deriving and applying water vapor corrections for GHGs and trace gases, especially in a field setting, over time. This approach seems like a viable approach for field deployment.
- The clearest benefit to the community is in communicating the design of the apparatus and experiment and in evaluating its efficacy. However, much of the paper reads as an assessment of a small number of instruments with a limited number of experiments. It seems to me that these assessments provide an incomplete picture of the instruments tested, and sometimes leave more questions open than they answer. Focusing the paper on the performance of the apparatus (as opposed to the instruments under test) would seem to be of greater benefit to the community, and better reflect the purpose of the paper as indicated by the title.
- In the assessment of the apparatus, I see a few gaps that I think merit additional testing and discussion:
- A better assessment of the performance of the water vapor generator would be helpful: accuracy, linearity, repeatability / stability over time and over ambient temperature, etc.
- The authors use the instruments themselves as water vapor monitors. Did they ever measure the same water vapor sample with two or more instruments to compare the measurements of water vapor?
- The apparatus relies on the fact that the nafion membranes do not allow significant quantities of the analyte gas to pass through the membrane. As far as I’m aware, the literature evidence shows small effects on CO2 and CH4 when there is gas on both sides of the membrane. Is there any literature on immersing the nafion in liquid? It would be very helpful if the authors performed an experiment where they humidified the calibration standard with the apparatus and then dried it, to demonstrate the accuracy of the humidifier apparatus. Without this test, there could be a bias in the humidity correction depending on the dissolved gas concentrations in the water.
- The authors are inconsistent in their choice of units for water vapor concentration. Please select one and use it consistently.
Specific Comments
L44: please provide reference for WMO/GAW compatibility standards.
S2.1: Apparatus. This is an interesting design. It looks as though there is a possibility of condensation occurring in the wet side output lines – they may need to be heated for better robustness.
I would like to see more testing just of the setup itself, with just one of the instruments and gases. Do you get the same answer when you scan the apparatus up in concentration and then back down? Does the step size matter? Do the steps need to be in order of increasing or decreasing concentration, or can they be randomized? Is there benefit to keeping the flow through water chamber fixed, or was it varied? Can an effect be seen in the resulting H2O and or analyte concentrations?
L132: it would be helpful to describe how the offset was included in Eq. 1
L153-L155: the selection of RS is not a particularly helpful statistical choice. For example, for a highly linear dependence (which would presumably be a desirable outcome), the RS of the quadratic term will tend to infinity. Perhaps just estimating the error in the correction from each term at a couple different water vapor concentration levels would be more clear and helpful, or reporting the std dev of each coefficient (not normalized by the mean)
Figure 3: (also applies to Fig 5,6,7): Putting the wet/dry ratio and the residuals in the same panel is very unclear. Please split these into two panels, where the top panel is the wet/dry ratio and fit, and the bottom panel is the residual of the fit. The symbols in general are quite small and difficult to read.
Figure 3: it would be nice to show the time series associated with one of these plots.
Fig 3: There are two sets of tests, one from 0 – 3% water, and one from 0 – 2500 ppm water. It is confusing that the polynomial fit over the wide range also includes the low range. If the authors are suggesting a piecewise fit, then I’d like to see clearly how they operate in the transition region. Otherwise, it’s difficult to know what correction to use in which situation.
L276: The authors note several limitations of this study. Of them, I think that the short period of time, and the small number of tests, is the most consequential, and would in principle be the easiest to remedy with minimal extra effort. For example, I would love to see repeated tests over a period of 1-4 weeks, showing e.g. hourly measurements of the wet-dry curve for one instrument and one gas, showing setup repeatability, drift over time, as water levels drop.
Citation: https://doi.org/10.5194/egusphere-2026-3732-RC1 -
AC1: 'Reply on RC1', Yunsong Liu, 17 Oct 2026
We would like to thank the referee for the review and comments to improve the quality of this work. Following these comments, we have made major modifications and point-by-point responses. Please find our response below (in blue).
General Comments.
- There is a clear need for deriving and applying water vapor corrections for GHGs and trace gases, especially in a field setting, over time. This approach seems like a viable approach for field deployment.
- The clearest benefit to the community is in communicating the design of the apparatus and experiment and in evaluating its efficacy. However, much of the paper reads as an assessment of a small number of instruments with a limited number of experiments. It seems to me that these assessments provide an incomplete picture of the instruments tested, and sometimes leave more questions open than they answer. Focusing the paper on the performance of the apparatus (as opposed to the instruments under test) would seem to be of greater benefit to the community, and better reflect the purpose of the paper as indicated by the title.
- In the assessment of the apparatus, I see a few gaps that I think merit additional testing and discussion:
- A better assessment of the performance of the water vapor generator would be helpful: accuracy, linearity, repeatability / stability over time and over ambient temperature, etc.
We agree that we have not addressed the accuracy of the water vapor generator in that we have not measured the water vapor mole fractions with an independent instrument. We unfortunately do not have access to a chilled mirror hygrometer or similar. For our application, the exact water vapor mole fraction isn’t crucial as – we just want to correct the trace gas measurements for the effects of water vapor, – but for other applications the accuracy would absolutely be important. We thank the reviewer for pointing out this oversight on our part and we have corrected the text in Lines 382-388 to acknowledge this limitation of our study.
We’re not sure what the reviewer means by linearity.
For our application, the key considerations are the repeatability and temporal stability of the water vapor calibration curves. In Table 1 and 2 and the original Figures 3-10, we included a total of 8 tests for A778, 10 tests for A798, 5 tests for CFKADS2025, 4 tests for CFKADS2218, and 5 tests for CKADS2025. The Aeris tests spanned 3-4 weeks for each instrument, whereas the Picarro tests spanned only 1-2 days for CFKADS2218 and CKADS2025, but approximately 3 weeks for CFKADS2025.
To more fully assess the repeatability of the results, we analyzed the saturation test results in this context for A798 (figure below). One test was performed on 17 March 2026, followed by five consecutive tests on 14 April 2026. At high water vapor concentrations, the resulting CH4 correction differed by 4.7 ppb (the March 17 correction minus the mean of the April 14 corrections at 20,000 ppm H2O). The differences were smaller at lower water vapor concentrations, with differences of 0.1, 0.5, and 1.5 ppb CH4 at 1500, 5000 and 10000 ppm H2O, respectively. Thus, the repeatability is better than the WMO target compatibility of 2 ppb for H2O less than 10000 ppm, but we do note that the full calibration curve is not adequately described by a second order polynomial for this instrument and the residuals are large for very dry conditions (below 2000 ppm H2O).
We performed the same analysis for the low water vapor tests listed in Table 1, which were conducted on 16-18 March. At 1500 ppm H2O, which was within the measured range of all four tests, the fitted CH4 correction due to water vapor ranged from approximately -2.0 to 0.6 ppb. This plot more clearly represents the repeatability of the calibration curves, so we have added this plot in the supplement (Figure S6) and the analysis (Lines 264-271) to more explicitly characterize the repeatability of the water vapor correction.
Figure S6. (a) CH4 wet-response repeatability of the Aeris MIRA serial number A798 analyzer over 28 days for the saturation tests. The y axis shows the CH4 correction, defined as the difference between the wet and corrected CH4 mole fractions (CH4_wet – CH4_corrected). Different colors indicate saturation tests S1-S6. Lines represent the fitted water-vapor response functions for each range. (b) CH4 wet-response repeatability of the Aeris MIRA serial number A798 analyzer over 3 days for the low water vapor tests.
Thank you for also bringing up the stability of the results over temperature. If the water generator is brought to a field site, it is likely that the temperature is not as well controlled as our laboratory. Of course, the ambient temperature itself determines the saturation water vapor level, so performing the testing in an unheated shed in the winter, for example, would not be ideal. This reviewer comment prompted us to examine whether temperature variability could account for the observed short-term H2O variability during the calibration steps for developing the calibration curve. We compared the standard deviation of the H2O during the calibration steps to the standard deviation of the enclosure inlet temperature (T2, see below). The correction coefficient (R2) was 0.006 for the 313 windows from the 3-5 min tests. Thus, variation in enclosure inlet temperature explains essentially none of the observed short-term H2O variability, which may be caused by Aeris H2O measurement noise, MFC flow-control variability, and/or analog setpoint-voltage stability. We added this figure to the supplement (Figure S5) and added the description text to Lines 257-262.
Figure S5. Standard deviation of the H2O during the calibration steps to the standard deviation of the enclosure inlet temperature (T2) of 21 tests (313 windows) for Aeris MIRA analyzers (serial number A778 and A798). The correction coefficient (R2) was 0.006 for these 3-5 min tests.
- The authors use the instruments themselves as water vapor monitors. Did they ever measure the same water vapor sample with two or more instruments to compare the measurements of water vapor?
The reviewer is corrected to point out that we do not have any head-to-head comparisons of our system to an existing humidity measurement (beyond that of the instruments to be calibrated). Using a sensor, such as a chilled mirror hygrometer inline would be a potentially valuable addition, particularly if the system is to be used for other applications. We added this point to the discussion (Lines 382-388). However, for our application, the key point is the value of water reported by the instrument, since, in practice, that is what we will be using in the field to correct for the remaining water vapor after using a Nafion dryer to remove most of the water. The actual value of the water and its calibration is not crucial (unless the calibration of the water reported by the instrument changes dramatically in the time between calibrations or during calibrations of during individual steps of the calibration).
- The apparatus relies on the fact that the Nafion membranes do not allow significant quantities of the analyte gas to pass through the membrane. As far as I’m aware, the literature evidence shows small effects on CO2 and CH4 when there is gas on both sides of the membrane. Is there any literature on immersing the Nafion in liquid?
As for previous literature, this concept has been used a couple times that we found. Dembowski et al. (2021) used this concept (with mass flow controllers and a submerged Nafion tube) to study the role of relative humidity in determining which aluminum mineral forms under various conditions. The NOAA system in Rella et al. (2013) used a gas-permeable membrane-based water calibration in a water bath, so not a Nafion but a similar idea. We addressed the permeation concerns with results shown in the next comment. In summary, the submerged Nafion did not show any effects of permeation, but we did see a small effect for a Nafion with gas on both sides inline after the water generator. Based on this, we do not believe permeation in our water generator to be a significant concern.
References
Dembowski, M., Loring, J. S., Bowden, M. E., Reynolds, J. G., Graham, T. R., Rosso, K. M., and Pearce, C. I.: The controlling role of atmosphere in dawsonite versus gibbsite precipitation from tetrahedral aluminate species, Dalton Trans., 50, 13438–13446, https://doi.org/10.1039/d1dt02081a, 2021.
Rella, C. W., Chen, H., Andrews, A. E., Filges, A., Gerbig, C., Hatakka, J., Karion, A., Miles, N. L., Richardson, S. J., Steinbacher, M., Sweeney, C., Wastine, B., and Zellweger, C.: High accuracy measurements of dry mole fractions of carbon dioxide and methane in humid air, Atmospheric Meas. Tech., 6, 837–860, https://doi.org/10.5194/amt-6-837-2013, 2013.
- It would be very helpful if the authors performed an experiment where they humidified the calibration standard with the apparatus and then dried it, to demonstrate the accuracy of the humidifier apparatus. Without this test, there could be a bias in the humidity correction depending on the dissolved gas concentrations in the water.
Thank you for raising this issue, prompting us to do additional testing. As shown in the figure below, a calibrated reference cylinder (424.4 ppm CO2 and 1977.2 ppb CH4) was measured with Picarro analyzer CFKADS2025 under three configurations: directly after the humidity generator, after the humidity generator followed by a downstream Nafion dryer, and through the Nafion dryer with the humidity generator bypassed. The external downstream Nafion dryer was a 144-inch Perma Pure Nafion tube (MD-070-144S-2), with counterflow provided by a zero cylinder (0.4 ppm CO2, 17.6 ppb CH4, 14 ppb CO). In (a) we see the water vapor, as measured by the Picarro during these three tests. In (b), the Generator only line indicates the CO2 changing by up to 0.175 ppm for water vapor of 1.4%. After drying with the Nafion, there is a possible permeation effect from the external Nafion of up to 0.075 ppm, essentially the same as when only the Nafion was used (no water generator). Since the Generator plus Nafion line is essentially the same as the Nafion only line, this indicates that the Generator itself did not affect the CO2 of the air sample. In (c), there is possibly a slight water vapor dependence for the Generator only test, but no evidence of permeation effects caused by the Generator or the external Nafion. In (d), we see a large water vapor effect for CO, and again no evidence of permeation effects. We have added this figure to the supplement (Figure S8) and text to the Results section (Lines 320-332).
Figure S8. (a) water vapor mole fraction measured by Picarro (CFKADS2025), for three tests (calibration cylinder gas through the water generator (Generator only), calibration cylinder gas through the water generator and an external Nafion (Generator + Nafion), and calibration cylinder gas through an external Nafion (Nafion only). (b) CO2 mole fraction (corrected to the dry mole fraction using the manufacturer correction). (c) CH4 mole fraction (corrected to the dry mole fraction using the manufacturer correction). (d) CO mole fraction.
- The authors are inconsistent in their choice of units for water vapor concentration. Please select one and use it consistently.
Thank you for pointing this out. We have changed the text and figures to consistently use ppm H2O.
Specific Comments
L44: please provide reference for WMO/GAW compatibility standards.
Thank you for the comment. The reference (WMO, 2025) has been added to Line 44.
References
WMO (2025): Measurement guidelines for the observations of carbon dioxide, methane, and nitrous oxide at global atmosphere watch stations, GAW Report No. 318, World Meteorological Organization, Geneva, Switzerland.
S2.1: Apparatus. This is an interesting design. It looks as though there is a possibility of condensation occurring in the wet side output lines – they may need to be heated for better robustness. I would like to see more testing just of the setup itself, with just one of the instruments and gases. Do you get the same answer when you scan the apparatus up in concentration and then back down? Does the step size matter? Do the steps need to be in order of increasing or decreasing concentration, or can they be randomized? Is there benefit to keeping the flow through water chamber fixed, or was it varied? Can an effect be seen in the resulting H2O and or analyte concentrations?
We thank the reviewer for these thoughtful suggestions regarding the robustness and operation of the P-WAVES apparatus. We agree that the response of the system to different water-vapor concentration sequences and step durations is important for evaluating its practical operation. We found that 3-5 min step produced very similar results, indicating that this duration was sufficient for the system and analyzer response to stabilize for the conditions tested. We also found that increasing water-vapor concentrations was preferable to decreasing concentrations. This behavior is likely related to the relatively slow removal of water from the Nafion tubing and other wetted surfaces (“stickiness” of water). When the water-vapor concentration was decreased, a longer equilibration period was required to reach the new water-vapor level. Consequently, concentration steps in the decreasing direction would require longer durations than 3-5 min steps used in our standard procedure.
We did not systematically randomize the concentration sequence because the same water-retention behavior makes randomized sequences difficult to implement efficiently: a transition from a high to a low water-vapor concentration can require substantially longer equilibration times. Thus, for routine measurements, we used an increasing sequence of water-vapor concentrations, which provided more rapid and reproducible equilibration. A randomized sequence could be used in principle, but longer equilibration periods would be required to ensure that the previous water-vapor condition did not influence the subsequent measurement.
During the experiments, the flow was kept constant at 160 sccm. We chose a fixed flow rate to maintain consistent operating conditions among concentration steps and experiments. The system could potentially accommodate a higher flow rate, but we did not systematically investigate the effect of flow rate on the generated humidity or analyzer response in this study.
Regarding condensation in the wet-side output lines, we recognize that condensation could potentially occur if the gas temperature falls below its dew point, particularly at high water-vapor concentrations. The present experiments were conducted without active heating of the output lines, and our results did not indicate an obvious condensation-related artifact under the conditions tested. Nevertheless, we agree that heating the wet-side lines could improve the robustness of the system. Particularly for operation at high humidity, large temperature gradients, or longer sampling lines. We have added this consideration to the discussion of the apparatus limitations (Lines 376-381).
L132: it would be helpful to describe how the offset was included in Eq. 1
We have clarified how the offset correction was incorporated into Eq. 1. The text “The offset was determined from the difference between the known values of the cylinder for each species and the uncorrected (wet) mole fractions measured by the analyzers at the lowest water vapor level.” was added to Lines 164-165.
L153-L155: the selection of RS is not a particularly helpful statistical choice. For example, for a highly linear dependence (which would presumably be a desirable outcome), the RS of the quadratic term will tend to infinity. Perhaps just estimating the error in the correction from each term at a couple different water vapor concentration levels would be more clear and helpful, or reporting the std dev of each coefficient (not normalized by the mean)
Thank you for the comment. We agree that the relative stability (RS) of the quadratic term can become very large when the underlying response is highly linear. We removed RS from the revised manuscript and instead report the standard deviation of each fitted coefficient across repeated measurements. We have updated Table 1 and 2, and revised the corresponding text.
Figure 3: (also applies to Fig 5,6,7): Putting the wet/dry ratio and the residuals in the same panel is very unclear. Please split these into two panels, where the top panel is the wet/dry ratio and fit, and the bottom panel is the residual of the fit. The symbols in general are quite small and difficult to read.
Thank you for the comment. Figure 3 and 5-7 were updated as the reviewer suggested.
Figure 3: it would be nice to show the time series associated with one of these plots.
Thank you for the comment. In the revised manuscript, the updated Fig.2 (a) shows the time series plot of one of the saturation tests performed with the Aeris MIRA A778. We believe this addition provides a clearer view of the temporal behavior of the measurements.
Fig 3: There are two sets of tests, one from 0-3% water, and one from 0-2500 ppm water. It is confusing that the polynomial fit over the wide range also includes the low range. If the authors are suggesting a piecewise fit, then I’d like to see clearly how they operate in the transition region. Otherwise, it’s difficult to know what correction to use in which situation.
Thank you for raising this issue and its potential confusion. We did not intend to suggest a piecewise fit. We suggest using the low water vapor fit if the sample air is dried (partially – there are still water vapor effects) and for the best compatibility using this approach for measuring greenhouse gases and associated tracers. If, for some reason drying is not desirable or practical, the correction to a wide range of water vapor values can be used. The analysis over a wide range also shows the expected errors if the manufacturer-corrected mole fractions are utilized, so that users can assess what works best for their application.
L276: The authors note several limitations of this study. Of them, I think that the short period of time, and the small number of tests, is the most consequential, and would in principle be the easiest to remedy with minimal extra effort. For example, I would love to see repeated tests over a period of 1-4 weeks, showing e.g. hourly measurements of the wet-dry curve for one instrument and one gas, showing setup repeatability, drift over time, as water levels drop.
Thank you for the comment. As stated for the previous comment about the repeatability and drift over time. We have added the analysis to Lines 264-271 and Figure S6 to the supplement.
The water reservoir contains approximately 130 ml of water and each test to saturation uses about 0.1 ml. Thus 50 tests to saturation thus use about 5% of the total volume of water so the reservoir level is essentially constant on the timescale or number of tests used here. We have added these details to the manuscript (Lines 99-101).
-
RC2: 'Comment on egusphere-2026-3732', Anonymous Referee #2, 07 Aug 2026
The comment was uploaded in the form of a supplement: https://egusphere.copernicus.org/preprints/2026/egusphere-2026-3732/egusphere-2026-3732-RC2-supplement.pdf
-
AC2: 'Reply on RC2', Yunsong Liu, 17 Oct 2026
We would like to thank the referee for the review and comments to improve the quality of this work. Following these comments, we have made major modifications and point-by-point responses. Please find our response below (in blue).
General Comments:
This manuscript details a low-cost, lightweight and portable humidification system for calibrating trace gas analyzers. In short, the system uses a combination of “wet” (produced by humidifying air through a Nafion membrane in a water bath) and “dry” (dry standard gas) air streams to produce stable water vapor levels comparable to ambient levels for trace gas analyzers in an automated fashion using pre-programmed Raspberry-Pi control. Currently, maximum water vapor levels achievable are dependent upon water bath temperature and mass flow of air through the system. The system is lightweight enough and has a low enough footprint that it can be efficiently moved from one observations site to another for deriving dry mole fractions of trace gases using empirically-derived water vapor corrections with this system.
The system was tested on both Aeris MIRA and Picarro CRDS systems, and indicates lower errors in corrected dry mole fraction measurements of trace gases relative to instrument-provided water vapor corrections, in addition to general adherence to recommended WMO compatibility goals for specific trace gases. The system is generally well-documented here and presents a worthwhile contribu9on to AMT over current methodologies; I provide some comments below to mainly clarify the description, potentially additional testing involved, and some additional details for consideration before publication.
Thank you for your time and expertise in reviewing and thus improving this paper. We appreciate your comments.
- Figure 1 and the main text description could use some expansion for clarity. First, please state whether there is any temperature control at all within the system at present. My understanding is that the water bath is not temperature-controlled. In L87, the stability of the water vapor levels is controlled by the stability of the water bath, so if this changes due to room temperature, etc., how does this affect the stability of the water calibration steps?
The reviewer is correct in that the system is not temperature-controlled at all and we have clarified this in the manuscript (Line 102). We note in the discussion that temperature control is a possible future improvement (Line 380). This reviewer comment prompted us to examine whether temperature variability could account for the observed short-term H2O variability during the calibration steps. We compared the standard deviation of the H2O during the calibration steps to the standard deviation of the enclosure inlet temperature (T2, see below). The correction coefficient (R2) was 0.006 for the 313 windows from the 3-5 min tests. Thus, variation in enclosure inlet temperature explains essentially none of the observed short-term H2O variability, which may be caused by Aeris H2O measurement noise, MFC flow-control variability, and/or analog mass flow controller setpoint-voltage stability. We added this figure to the supplement (Figure S5) and added the description text to Lines 257-262.
Figure S5. Standard deviation of the H2O during the calibration steps to the standard deviation of the enclosure inlet temperature (T2) of 21 tests (313 windows) for Aeris MIRA analyzers (serial number A778 and A798). The correction coefficient (R2) was 0.006 for these 3-5 min tests.
Please specify tubing composition – is this stainless steel (and which diameter)?
The system uses 1/8-inch outer-diameter (OD) stainless-steel tubing and we apologize for the oversight in not including this important detail. Added to the manuscript (Line 103).
Please describe the purpose of the overflow in Figure 1 for the reader (presumably for
analyzer pressure stability).
Thank you for the comment. The text “An overflow line was included downstream of the system to maintain a stable flow and pressure at the analyzer inlet.” was added to Lines 97-98.
- Figure 2: it is difficult to see in a) how stable the water actually us at the end of the 3-5 min step – can you zoom in on a few of these steps to show that the H2O mole fraction equilibrates relatively quickly to a stable value? As you describe tests with both the Picarro (up to 1.7% H2O) and the Aeris instrument later in the text (up to 3% H2O) and as it looks as if the precision of the water vapor mole fraction degrades at higher H2O, can you provide results in b) that substantiate your claim for stability and repeatability in the humidity control up to these H2O mole fractions? The example of only stepping up to 0.7% in Section 2.1 seemed somewhat incomplete given that the rest of the manuscript describes system tests to higher H2O levels.
Thank you for this comment. We added representative insets for two steps to Fig. 2. It is now more clear that the water vapor levels off quickly (see below – which is the updated Fig. 2). We also agree that we should show steps up to as high as possible. You can see that indeed the stability does degrade at the higher water vapor levels. In particular, Step 20 doesn’t appear to have leveled off at all. Previously we had focused our efforts on the lower water vapor levels because with our system in the field, we dry (albeit not completely, hence the need for water vapor calibration), and thus quoted 3 ppm H2O SD. But in fact, the stability is more like 10 ppm H2O for water vapor up to 3% and even higher above. We adjusted the text to reflect these nuances (Lines 135-141).
We also added a figure (below) to the supplement (Figure S1) to more carefully address the water vapor variability at various levels during the steps used to develop water calibration curves for the instruments. The variability increases with increasing water vapor for the completed tests. Across 29 humidity-step tests comprising 368 individual steps and two Aeris analyzers, the median final-60-s H2O variability was 2.8 ppm (1σ), with typical variability of ~2-3 ppm below 20,000 ppm H2O and ~7-10 ppm at the highest H2O levels.
Figure S1. Variability of generated H2O as a function of water vapor concentration and step duration. Each symbol represents the standard deviation of H2O during the final 60s of an individual concentration step. Different symbols indicate step durations of 3, 3.5, 4, and 5 min. The black line represents the binned median across all 3-5 min tests.
- Describe which MFCs would be needed to achieve higher water vapor levels comparable to those observed for the Picarro analyzer.
Thank you for the comment. We just need both the wet and dry MFCs to have a maximum flow rate slightly in excess of the flow that the instrument needs. For example, the Picarro draws about 250 sccm so each MFC needs to have a max control value of about 275 sccm, assuming we are using an overflow method.
- L294-295 states that the Nafion-based system should minimize solubility-related effects from the water bath since there is no direct air-to-gas interaction, and before this, that permeability effects are somewhat negligible. However, have you tested this for CO2 (and potentially also CH4) when the water bath is not acidified? It is still somewhat unclear to me whether deionized or more basic water baths could still create issues with the Nafion system at higher water vapor levels > 1.5%.
Thank you for raising this issue, prompting us to do additional testing. As shown in the figure below, a calibrated reference cylinder (424.4 ppm CO2 and 1977.2 ppb CH4) was measured with Picarro analyzer CFKADS2025 under three configurations: directly after the humidity generator, after the humidity generator followed by a downstream Nafion dryer, and through the Nafion dryer with the humidity generator bypassed. The external downstream Nafion dryer was a 144-inch Perma Pure Nafion tube (MD-070-144S-2), with counterflow provided by a zero cylinder (0.4 ppm CO2, 17.6 ppb CH4, 14 ppb CO). In (a) we see the water vapor, as measured by the Picarro during these three tests. In (b), the Generator only line indicates the CO2 changing by up to 0.175 ppm for water vapor of 1.4%. After drying with the Nafion, there is a possible permeation effect from the external Nafion of up to 0.075 ppm, essentially the same as when only the Nafion was used (no water generator). Since the Generator plus Nafion line is essentially the same as the Nafion only line, this indicates that the Generator itself did not affect the CO2 of the air sample. In (c), there is possibly a slight water vapor dependence for the Generator only test, but no evidence of permeation effects caused by the Generator or the external Nafion. In (d), we see a large water vapor effect for CO, and again no evidence of permeation effects. We have added this figure to the supplement (Figure S8) and text to the Results section (Lines 320-332).
Figure S8. (a) water vapor mole fraction measured by Picarro (CFKADS2025), for three tests (calibration cylinder gas through the water generator (Generator only), calibration cylinder gas through the water generator and an external Nafion (Generator + Nafion), and calibration cylinder gas through an external Nafion (Nafion only). (b) CO2 mole fraction (corrected to the dry mole fraction using the manufacturer correction). (c) CH4 mole fraction (corrected to the dry mole fraction using the manufacturer correction). (d) CO mole fraction.
Specific Comments:
- Title: you might consider renaming to something more general like “Portable system for characterizing trace gas analyzers for sensitivity to water vapor” as you describe water vapor corrections for CO in the main text.
This is a good idea, thank you for the suggestion. “Portable system for characterizing trace gas analyzers for sensitivity to water vapor: Application to cavity ring down and mid IR sensors” to emphasize that our purpose is the application.
- L44: Please provide a citation for the stated WMO compatibility goals. These are often
found in WMO annual meeting reports.
Thank you for the comment. The reference (WMO, 2025) has been added to Line 44.
References
WMO (2025): Measurement guidelines for the observations of carbon dioxide, methane, and nitrous oxide at global atmosphere watch stations, GAW Report No. 318, World Meteorological Organization, Geneva, Switzerland.
- L47: “500 ppm H2O potentially resulting in significant measurement errors” – Rella et al. 2013 state that there are CO2 measurement biases from 500 ppm H2O, so you could cite this paper here for reference.
Thank you for the comment. The reference (Rella et al., 2013) has been added to Line 48.
- L90: Specify maximum and/or average power draw to better substantiate the potential
for portable operation under battery power.
Reviewer CC1 also requested this, which was an overcite on our part. Thank you for pointing this out. The text “During normal operation, the P-WAVES system consumes approximately 10 W of electrical power. A 24 h deployment would therefore require approximately 240 Wh of energy. As a representative example, a commercially available 12 V, 30 Ah LiFePO4 battery has a nominal capacity of approximately 360 Wh and could provide sufficient energy for more than 24 h of operation, depending on power-conversion efficiency and usable battery capacity. A representative battery of this size measures approximately 16.5 × 16.3 ×8.6 cm, weighs about 3.2 kg, and could be accommodated within the enclosure used for the current P-WAVES configuration.” was added to Lines 112-117.
- L102 “…requires longer steps” – Do you have a sense for how long?
Thank you for the comment. The text “(about 10 min)” was added to Line 123.
- L139: Please also state value of CO dry mole fraction for completeness.
Thank you for the comment. The text “For CO, the value of the calibration cylinder was estimated to be 123.8 ppb.” was added to Line 173.
- L142: It would perhaps be useful to state how long it takes for the H2O mole fraction to equilibrate (also see comment for Figure 2 above) to provide justification for 3-5 minute time steps.
Thank you for the comment. As mentioned for the previous comment, Figure 2 has been updated and Figure S1 was added to supplement. The text “The short-term within-step stability gradually decreased at higher water vapor levels. Across 29 humidity-step tests comprising 368 individual steps and two Aeris analyzers, the median final-60-s H2O variability was 2.8 ppm (1σ), with typical variability of ~2-3 ppm below 20,000 ppm H2O and ~7-10 ppm at the highest H2O levels (Fig. S1).” was added to Lines 139-141.
- L215: It is somewhat surprising that there was no water vapor dependence for CO for
the CKADS2025 as these analyzers in the past have shown a water vapor dependence
(Chen et al., 2013). You may note or comment for the reader that this is somewhat unexpected.
As we noted on Page 12, Lines 241-243, two of the Picarros indicated significant CO dependence but CKADS2025 did not. The text “For these instruments, the measured CO mole fraction is commonly observed to depend on water vapor (Chen et al., 2013).” was added to Lines 311-312.
- Figure 8: It may be useful to provide the same y-axis scale for Figures 8(a-c) to compare variability between tests.
Thank you for the comment. Figure 8 was updated as suggested.
- L338: “if a linear dependence is assumed” – but we know that the CO bias is non-linear (Chen et al., 2013); perhaps caveat this statement in some way?
Thank you for the comment. The sentences were rewritten as “This study showed that without analyzer-specific water correction for CO, the mean CO bias reached about 12 ppb for the CFKADS2025 and CFKADS2218 analyzers for up to 17000 ppm H2O. A linear extrapolation would yield an estimated bias of about 22 ppb at 30000 ppm H2O, however, this extrapolation should be interpreted cautiously because the CO response to water vapor is nonlinear (Chen et al., 2013).”
Technical Comments:
- L41 and others: please list citations in ascending order by year
Thank you for the comment. We have revised the citations throughout the manuscript to list them in ascending order by year.
- L82: please define acronym ‘MFC’ here as it is used in Figure 1 below.
Thank you for the comment. We have defined the acronym “MFC” at Line 82.
- L128 – please punctuate the end of this sentence with a period after Eq. 1.
Thank you for the comment. We have added a period after Eq.1 as suggested.
- Figure and Table appearance: For Figs 3-10 in general, it is difficult to differentiate
between the triangles and circles for different dry vs. corrected values. One suggestion
may be to use filled vs unfilled shapes for dry vs. corrected values. Shapes in Figures 5-10 are somewhat small, as is the text in Table 2 (maybe this can be rotated 90 degrees).
Thank you for the comment. Figure 3-10 and Table 2 were updated.
- L241: “Picaro” Picarro
Thank you for the comment. “Picaro” was corrected to “Picarro” at Line 241.
-
AC2: 'Reply on RC2', Yunsong Liu, 17 Oct 2026
-
EC1: 'Comment on egusphere-2026-3732', Thomas Röckmann, 20 Aug 2026
Dear authors,
As you have seen, the expert referees have brought up quite some concerns regarding both the novelty of the approach and the rigorous of the testing of the system. Important aspects are the statistical evaluation, precise experimental description, calibration, and the short-and long term stability of the system described. These are important points that will require thorough major corrections to the manuscript and likely also additional experiments as indicated by the expert referees. If you think you are able to assess these points, you are welcome to submit a thoroughly revised version, which will then be sent again to the referees for evaluation.
Best regards
Thomas Röckmann
Citation: https://doi.org/10.5194/egusphere-2026-3732-EC1
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 334 | 160 | 48 | 542 | 72 | 33 | 42 |
- HTML: 334
- PDF: 160
- XML: 48
- Total: 542
- Supplement: 72
- BibTeX: 33
- EndNote: 42
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1
The manuscript “Portable system for characterizing greenhouse gas analyzers for sensitivity to water vapor” by Liu et al. claims to describe a “lightweight, low-cost, and portable system” for performing water vapor corrections of greenhouse gas analyzers.
The basic premise of their system is forming a “wet” stream of calibration gas using a Nafion-based humidity exchanger, then combining it with a “dry” stream at set ratios using mass flow controllers and valves. This is a description of common laboratory practice that has been used for decades in air pollution / air quality and VOC laboratories and in instrument development and testing, in commercially available instruments (e.g. 2B Technologies DewLine), and presumably also used already in the greenhouse gas community.
The claim to novelty, then, comes from the weight, cost, portability, and apparent stability of their system. They provide dimensions (30.5 cm x 30.5 cm x 20.3 cm, 5 kg) but it is not clear whether that includes all parts of the system (including the water and a battery) or not. They claim “low power requirements” but do not specifically test or report the power draw or make any claims for how long a typical battery-operated system could run in the field. They also provide no head-to-head comparisons of their system to existing, commercially available systems to provide an equivalent comparison.
The authors make a claim of “high stability” (3 ppmv H2O). However, in the manuscript they appear to only take the average of the last 60 seconds of a 3.5 minute step. It would be much more convincing if they reported a response time or equilibration time to humidity change and provided the long-term (e.g. 8 hour, 24 hour, etc.) stability of the system instead of 60 second transients. A 60-second transient does not show “high stability”. Depending upon the type of tubing used (polymer, stainless steel) and specific instruments used, the system response time to water vapor could be on the order of 60 seconds or longer. The apparent high stability could be a system-induced broadening kernel. The authors do not state what materials, tube diameters and lengths, etc. they used, nor do they provide any residence time distribution data to prove that their 60-second stability is actually the stability of their humidity generation and not a system artifact.
The authors claim “stable and repeatable humidity control…” but provide no statistics or demonstration of long-term stability and either short-term or long-term repeatability. As far as I can tell, their experiments were purely done in a climate-controlled laboratory environment. This performance might not transfer to field environments. In particular, the stability of the humidity generation will be highly sensitive to the temperature of the system, so ambient temperature fluctuations will adversely impact the stability of the system. It is not clear that the authors have thoroughly evaluated and addressed these impacts.
The authors do not use any independent humidity or dewpoint sensors (such as a chilled mirror or capacitive sensor) to confirm the accuracy or stability of their humidity generation. Instead, they rely on instrument-reported water vapor lines. However, they provide no calibration or traceability for these measurements. As a humidification system or humidity generator it has no inherent humidity measurement capability.
As mentioned earlier, much of the system (split flows, Nafion humidification, mass flow controllers to adjust humidity, etc.) are already common practice in the field. If the authors want to make a novel claim about the superiority of their system they need to provide additional details about the other factors (mechanical designs, CAD files, Python code, parts lists, etc.) to make it reproducible and independently testable. If the authors can address some of these concerns and provide more convincing data about the value of their system I think it would be a useful contribution to the field, especially if the technology is made available (public domain or commercially).
I do not address the Aeris and Picarro testing here, as those aspects are already well-covered in the scientific literature and well known to be dependent upon instrument, firmware / software version, etc. and not necessarily transferrable (although the calibration process and applied correction models could be).