The manuscript presents an extension of previous work (Norooz Oliaee et al., 2022), advancing from instrument development toward field validation of UAV‑based CH₄ flux estimation using controlled‑release experiments. While the spectrometer design largely follows the earlier publication, the field deployment lacks a sufficiently rigorous assessment of the method under realistic conditions. In particular, the study does not include a systematic evaluation or quantification of uncertainties, biases, and errors related to interpolation approaches, instrumental performance, or atmospheric transport processes. With only two flights (of which just one was conducted under suitable wind conditions) the study does not provide valuable or generalizable insights for the reader.

The paper may be suitable for publication after the authors address the concerns listed below.

General comments:

Validation of plume‑mapping and flux‑estimation methods using sources with known emission rates has been extensively discussed in recent literature. Several studies (e.g., Andersen et al., 2021; Morales et al., 2022) present robust flight protocols and explicitly make clear suggestions for minimum wind speeds and other limiting factors that must be considered for appropriate quantitative flux analysis. These studies generally indicate optimal mean wind speeds of 2–6 m s⁻¹, while wind speeds <2 m s⁻¹ lead to unreliable flux estimates due to large plume variability. The authors are encouraged to review this literature and use these recommendations into the design of future field campaigns.

Existing flux‑estimation methods, including mass‑balance approaches (e.g., Morales et al., 2022) and Gaussian plume inversion modelling (e.g., Shah et al., 2020), typically show uncertainties ranging from ~17% to >200%, dominated by environmental conditions that control plume dispersion. Yet the present manuscript attributes its dominant uncertainty solely to the extrapolation between the lowest flight transect and the ground. It is unclear why this assumption was not verified using a ground‑based instrument (e.g., MIRA Pico Aeris) across the downwind plume (e.g. installed on a mobile platform such as a car).

A more detailed discussion on the accuracy of UAV‑based wind measurements is required, given that wind is a critical parameter in mass-balance calculations. The calibration and performance assessment of the spectrometer were carried out in the laboratory, which may not reflect real‑world conditions. The authors should clarify potential impacts of environmental perturbations: changes in airflow through the MPC (due to UAV speed, wind direction, hovering), variations in pressure and temperature, and the effect of these factors on the non‑linear calibration curve. Similarly, the potential dependence of measured fluxes on the true emission rate warrants discussion.

The manuscript does not sufficiently address the uncertainties introduced by the necessary interpolation across the vertical plume curtain. A more substantial dataset would be required to evaluate these effects robustly, which is not achievable within the reported 54‑minute field campaign.

The authors should also reconsider the title and abstract. The introduction suggests a focus on instrumentation, but the bulk text emphasizes flux estimation. Given that no substantive new instrumental development is presented, this contrast weakens the paper’s positioning. Furthermore, mass‑balance flux estimation involves more than “just” UAV‑based concentration measurements.

Abstract: Technical details of the data acquisition system are not central to the key findings and can be removed. The authors should instead summarize the field‑experiment results. The statement that the instrument “can be deployed to measure natural sources that would be undetectable for other methods” is too strong given that only a single suitable flight was analyzed. The authors should instead outline a realistic outlook, such as: "The instrument could be deployed in future controlled release experiments to validate its performance for detecting fugitive anthropogenic methane emissions (e.g., oil and gas infrastructure).”

Methods: 

The rationale for selecting WMS over DAS is not explained. The authors should briefly discuss the limiting factors for DAS in this context.

The manuscript should clarify what quantity the analyzer measures. Greenhouse gas concentrations are typically reported as dry mole fractions—how was water‑vapor effect corrected? Although the analyzer records H₂O, this was not shown or discussed.

Flux estimation requires interpolation of measurements along the vertical direction. A description of the interpolation method is missing.

Specific Comments

L90: Remove “free space” before “laser”.

L91: Specify the tuning rate used for the spectral scan. In Fig. 2c, replace channel number with optical frequency (cm⁻¹).

L96/98: Remove unnecessary trailing zeros.

L106: Remove “has no moving part”.

L110: Compare the 1.2 kg sensor package with the MIRA Pico regarding weight, precision, stability, and response time.

L117: Discuss the impact of rapid ambient‑temperature changes on detector responsivity.

L130: The timing‑accuracy statement is redundant, given that bandwidth was already defined.

L137: Clarify why a modulation current of 15 mA was chosen. The modulation index appears sub‑optimal in Fig. 2c.

L237: Remove “per minute” and “an in line”.

L245: Discuss the sensor response to sudden changes in pressure/flow/temperature.

L249: The calibration uncertainties (e.g., ~0.45 ppm at 2.5 ppm CH₄) are substantial. More detailed assessment at ambient CH₄ levels is needed.

L259: Additional in‑flight noise is attributed to inductive effects. How does this scale without real‑time data streaming? How can it be mitigated? Is this typical for UAV‑mounted spectrometers?

L269: Rephrase “flat region in the laboratory Allan measurements”.

L270: Specify the hardware and software filters used, including 3 dB roll‑off frequencies.

L271: Reconsider the wording of  “limited or flattened noise”.

Fig. 4: Strong filtering is observed here but not in Norooz Oliaee et al. (2022, Fig. 6). The reported 1‑s precision is also four times worse despite filtering. This discrepancy needs discussion.

L280: To avoid confusion, use consistent units (L min⁻¹ instead of introducing SLPM).

L294: The UAV wind speed is sampled at 2 Hz while concentration data are at 10 Hz. Explain the motivation for these differing sampling rates.

L305: Clarify what “very low” refers to regarding the 0.5 kg h⁻¹ emission rate.

References:

Andersen, T., Vinkovic, K., Vries, M.d., Kers,B., Necki, J., Swolkien, J., Roiger, A., Peters, W., Chen, H.: Quantifying methane emissions from coal mining ventilation shafts using an unmanned aerial vehicle (UAV)-based active AirCore system, Atmos. Environ, https://doi.org/10.1016/j.aeaoa.2021.100135, 2021.

Morales, R., Ravelid, J., Vinkovic, K., Korben, P., Tuzson, B., Emmenegger, L., Chen, H., Schmidt, M., Humbel, S., Brunner, D., 2022. Controlled-release experiment to investigate uncertainties in UAV-based emission quantification for methane point sources. Atmos. Meas. Tech. 15, 2177–2198. https://doi.org/10.5194/amt-15-2177-2022.

The manuscript entitles “Ultra-Lightweight Mid-IR Methane Sensor for UAV-based Measurements” by Beattie et al. presents an absorption spectrometer designed for airborne in-situ measurements on a small UAV platform. The study describes the optical and technical design of the sensor for methane detection and evaluates its instrumental performance through laboratory and field experiments. In addition, two field missions are presented in which methane plumes from controlled-release point sources are measured, and corresponding fluxes are derived using a curtain flux approach.

This manuscript is well written and presents an integrated work of instrument development and characterization to first field missions; however, it lacks of detailed method descriptions, instrument uncertainty characterizations, and analytical tests.

After major revision, this manuscript might be suitable for publication.

General comments:

The authors use throughout the whole manuscript the wording “concentrations”, even though the data is shown in mixing ratios (ppm or ppb). Mixing ratios are converted into concentration for the flux estimation, as also described in Section 2.7; however, the detection is via mixing ratios. The authors should not confuse this and stick to mixing ratios, when they report the values in ppm or ppb.

The manuscript would benefit from including more literature, especially in the introduction and the discussions. This sensor is not the first small and lightweight airborne absorption spectrometer (e.g., D’Amato et al., 2025). The authors should consider to compare their sensor to similar instrumentation, highlighting the specific advantages and limitations of the presented instrument relative to existing systems.

The authors should include more discussion on the instrument’s performance regarding uncertainty contributions. Literature has shown that laboratory performance cannot be assumed for field operation, particularly for airborne platforms (Werle et al., 1993, 2011, Röder et al., 2024, Ort et al., 2024). The authors already included the Allen-Werle plot for both, laboratory and field operation; however, optimum integration times and precisions are stated, but not discussed regarding their error contributions. Please state somewhere, what you mean with “ambient methane values” and how you defined those during both timeseries used for the Allan deviation. Were laboratory tests made to quantify the effects of pressure, temperature, and water vapor on the signal for typical ambient mixing ratios (near 2ppm)? Moreover, the Allan-Werle plot for the in-flight operation uses ambient measurements; how do the authors assure that there are no atmospheric changes in mixing ratio during this UAV flight period? Was this timeseries measured during similar movements of the UAV than during the A and B field missions, or was it at a fixed location? How was the wind velocity, pressure, etc.? Particularly in mid-infrared absorption spectroscopy, frequent fringe or etalon structures contribute to the signal uncertainty. How sensitive is the sensor to those error contributions, along with contributions from changing pressure, temperature, and water vapor? How do the authors justify to use a 1 Hz time resolution, where they only are able to detect atmospheric methane variabilities with a precision of 26 ppb. Moreover, total measurement uncertainty can not only be explained by the precision, particularly when the standard deviations of the timeseries in the lab and during flight in Figure 4a,b seems to be at +-50 ppb and +-250 ppb, respectively. According to the Allan-Werle deviation at 100 Hz resolution, this cannot fully be explained by precisions of approx. 8 ppb, and 70 ppb, for lab and in-flight respectively. How do the timeseries look like with the optimum averaging times? What is the total uncertainty predicted for the ambient observations, e.g., including drift, reproducibility, etc.?

The instrument performance and possible atmospheric applications need to be put into better context, and clarified, for which type of atmospheric methane sources the sensor can be used for (e.g., strong emitting sources such as oil/gas lacks, wastewater facilities (report corresponding literature)). I miss the statement how this controlled-released field campaigns are satisfied to represent real-emission scenarios. While the intentional release of a potent greenhouse gas at fluxes of approximately 0.5 kg h⁻¹ should be considered carefully, the study would benefit from an additional in-field characterization experiment. In particular, a comparison with simultaneous measurements from an independent ground-based instrument (e.g., Aeris Pico, MIRA Pico deployed on a tower or mobile platform) at a real methane emission source would help to better quantify the performance of the airborne sensor. Such an experiment would also provide a clearer demonstration of the system’s potential applicability for future in-field measurement campaigns.

The spectra in Figure 2 shows the absorption lines of CH4 and H2O for roughly 2ppm of CH4, right? The 2f spectrum is usually ideal to define a background line, with usually no absorption if there is no error contribution. If I understood it correctly, methane mixing ratios are computed by a peak-to-peak amplitude searching the minimum and the maximum signal within predefined bounds. That would work if the background line would not experience etalon or other sinusoidal structures, which it seems to experience, as it looks like in Figure 2c, and the Allan-Werle plot. Those structures would non-linearly expand (or decrease) the peak-to-peak amplitude, resulting in much higher (or lower) mixing ratios than existing. Have the authors tried to use an average background line instead of a minimum peak? What are the predefined bounds to compute the peak-to-peak amplitude? How much does the signal drift during operation, as no line-locking or calibration is done to identify the correct absorption line? How can it be assured that the H2O line does not lie within the predefined bounds at some point?

Throughout the manuscript several different time resolutions, and precisions are reported. For example, in the Abstract, the authors state a precision of 3.7 ppb at 1 s time resolution; but in the field missions they use an averaging time of 0.1 s with a reported precision of 26 ppb. Please decide which resolution and precision is appropriate for the sensor and assure continuity throughout the manuscript.

In-line comments:

Abstract:

• In the abstract, the key findings of this manuscript should be briefly reported and motivated. What is the motivation? What are the key aspects of the sensor (not necessarily need to report the electrical power)? And what test have been done to evaluate the instrumental performance?
• Line 1: change “mobile” into “in-situ”
• Lines 8-9: “undetectable for other methods”? This is a strong statement, and not true. Balloon-based or flux-tower based measurements, e.g., also provide such low-altitude in-situ data.

Introduction:

• The introduction generally needs more literature citations. There are a lot of facts stated without referring literature to it (e.g., line 18, 24, 30).
• Line 12: Change “processes” into “sources”
• Line 13-16: What about flux towers?
• Add a comma after “e.g.”
• Line 24-25: “>100 m” or “hundreds of meters”? It really depends here on the terrain and aircraft. Over the ocean a small aircraft can fly even lower than 100m. Maybe it would be better to emphasize the requirements of extrapolation, if difficult terrain and vegetation are present.
• Line 28: “greater distances” are also a huge benefit of aircraft observations, as those allow air parcel tracking or convective plume tracking for smaller up to longer distances (see Karion et al., 2015, Conley et al., 2017, Riese et al., 2025, Curtius et al., 2024)
• Line 30 – 31: “extremely stable and sensitive instruments” also exist particularly built for aircraft in-situ observations (see, e.g., D’Amato et al., 2025, Ort et al., 2024, Müller et al., 2015, Viciani et al., 2018)
• Lines 41-55: This comparison between “in-situ” and “ex-situ” (while the phrasing “remote-sensed” is more commonly used in atmospheric measurement techniques) is a bit out of context here. I see the point of motivating in-situ over remote-sensed techniques here, but maybe shortening this part and may expanding on comparisons between more similar drone sensors instead would highlight the benefits of the reported sensor more.
• Line 58: “a delayed response”: this depends on the pump and flow rate used in the system. Clarify more what you mean, add more citation.
• Line 62-64: It may be helpful here to clarify more which limitations and challenges exist for open-path systems and how your system is trying to tackle those.
• Line 65: weird sentence structure: “deployed for deployment”
• Line 67 (& line 84): Where do you define v₃? Do you need it anywhere again?
• Line 65-75: Sounds more like a conclusion. Maybe combining it with the part in line 76-80 and shorten it would help.
• Line 71 & line 77: “SBC” is defined twice here. Once is enough.

Section 2.1: This section feels unnecessary, as it repeats your last part in the introduction and detailed descriptions to the principles that are explained anyhow in the following Sections. I would remove this part, include some of it into later sections, or motivate in the introduction. For example, why haven’t you used DAS instead of WMS? Why TDLAS and not QCLAS, which is anyhow claimed to be more suitable for small and compact absorption spectrometers? This should be motivated more in the introduction.

Section 2.2: Include the temperature range of the laser

• Line 90: Write “Figure”, as it starts the sentence
• Line 92: What kind of Herriott cell exactly (spot-ring, astigmatic, etc.)? Add citation, accordingly.
• Line 94, 95: Remove the brackets at the citation
• Line 95: What is the reflectivity of the gold-coated mirrors? Please add.
• Line 98: I think the authors confused the unit of the path length here. Within a multi-pass cell of length 131.4 mm and 35 passes, 4.6 km path length sounds a bit too much for me.

Section 2.3: Move Section 2.3. after Section 2.5. This would improve the flow as you would have explained all parts already, which are mentioned here.

• Line 106: The optical alignment should be transferred into the Section “Optical assembly” and the alignment method should be described in more detail. How can shims achieve such fine injection angles? How was the correct adjustment approved? Did you use a visible light source for the adjustment? Was the cell itself already adjusted?
• Line 109: Clarify here again in brackets, which quantities the ambient condition sensors can measure.

Section 2.4:  Explain how the cooling of the laser and detector works, and what the heatsinks are there for. How much varies the laser and detector temperature in field operation? And what’s the range of impact on the signal with changing temperature?

Section 2.5:

• Line 135: Add citation of the HITRAN line for the chosen methane absorption line.
• Line 141: Why 2f and not, e.g., direct absorption? Explain more the reasons for your choice. Include references of airborne instruments also using 2f (ALIAS, BLISS, DACOM, ATLAS, ATTILA).
• Line 148: “section” to “Section”

Section 2.6:

• Clarify better how long the UAV can be operated in field. Line 166 and line 184 report two different (maximum) flight times. Stick here to what is possible for full operation mode under ambient conditions, considering wind and temperature ranges. Adjust the maximum operation time also in the conclusions (line 355).
• What is the maximum altitude this system can reach?
• Line 170: For both, vertical upward and downward operation? Are eddy-induced changes in wind velocity considered for downward flights?
• Line 183: change “cames” to “comes”

Section 2.7:

• Line 187: “flux curtain” pattern usually perform upwind flights as well, so that the background can be defined, without any disturbance from the point source. It is not clearly stated how the background value was defined for the two field missions.

Section 3.1:

• This section does quantify the sensors performance briefly, with three different approaches, but they all need to be more precise in discussing the sensors performance and limitations. According to the Allan-Werle deviation a minimum deviation of 3.4 ppb can be achieved at 5 s integration time, representing the optimal averaging time before instrumental drift becomes dominant. Of course, with faster time resolution, measurements can capture more atmospheric variability, but the atmospheric variability still needs to be larger than the instrumental variability. Furthermore, a total uncertainty estimation on the measured mixing ratios is missing (as mentioned in the general comments).  
• Line 226: change “2 ppmv” to “ppm” and its mixing ratio not concentration!
• Line 228: change “concentration” into mixing ratio and remove v/v.
• Figure 2: Including the full H2O line would allow to track water vapor mixing ratios simultaneously, which may be useful for the water dilution effect correction for the wet methane mixing ratios. Although, depending on the phase of water during measurements, as droplets may cause issues. 
• Lines 232 – 253: The sensor calibration was performed with mixing ratios ranging from 2 – 50 ppm. It would be interesting how the fitting curve and the comparison with the commercial methane sensors look like for more data points close to atmospheric typical mixing ratios. Higher mixing ratios also mean a stronger signal. How sensitive is the sensor to measure differences which are more likely to occur in the atmosphere? How are the uncertainties for mixing ratios close to 2 ppm, for both sensors?
• Lines 269-271: What about other error contributions? What about drift, sinusoidal structures? Please expand on the error contribution discussion here. Literature from Werle et al., 1993, 2011 or Röder et al., 2024 may help here.
• Figure 4: Is the timeseries a, b measured in 100Hz or 10Hz?

Section 3.2:

• This section would benefit from a more real-world field experiment, maybe with a comparison to other methane sensors to show the instruments in-field performance.
• Lines 278-279: This sentence sounds confusing.
• Lines 289: How do you measure the background?
• Line 301: 2.1 ppm background level is quite high (see Lan et al., 2026). Are you sure that you have measured the background? Highest values at 8.2 ppm are also quite high. Is this realistic for “real” anthropogenic methane emission sources? Is there literature reporting similar values in the atmosphere from such anthropogenic sources?
• Lines 308-309: Why not? Either show the results and discuss them with literature why the controlled-release B did not work or not include this field mission B at all. It may be anyhow better to include here a direct comparison with another instrument in field close to, e.g., a real gas/oil leak to also show a possible application of this sensor and its advantages against other methane sensors.

Section 3.3:

• Line 314: Write always “Figure” instead of “Fig.” if it starts the sentence.
• Line 318: Is it defined somewhere how the calibration uncertainty is calculated?
• Line 319: The 26 ppb from the Allan-Werle plot represents the precision at integration times of 0.1s (see Werle et al., 1993).
• Line 320: Why is the background averaged over 60s? Explain why you chose this. And what do you mean with “resolution of 10 ppb”? Don’t you mean a “precision of 10 ppb”?
• Line 320: “Min” is not defined. Replace ms^-1 with ms^-1.
• Figure 6: Have you measured the vertical wind speed or is vertical advection only derived from the model?  
• Line 328: Not a complete sentence. Please repeat here briefly what the parameters mean for a better reading flow.
• Line 330: What is the reason for the delay of 23 minute (e.g., logistics, battery charging)? Were calibrations performed in between?
• Lines 332-333: Why is the far flux higher than the near flux? It falls within the uncertainty but could there be another explanation (e.g., wind, full plume advection into flight tracks only at far curtain, stronger uncertainties)? Particularly, as the mixing ratios were much higher in the near flux curtain.
• Line 338: Would be always beneficial to measure the whole vertical profile, but the plume propagation depends also a lot on local advection. Are there radiosonde profiles near by in upwind direction to somehow support this large flux underestimation without the extrapolation with the meteorological conditions?

Conclusion:

• The conclusion should be considered to be structured differently. What is the key point of this paper? Start with the general construction of the instrument, then report laboratory and field performance, followed by the field evaluation of the sensor with the example missions. Clearly state what are the advantages of this sensor, and what are possible applications this sensor can be used for. Also discuss briefly its limitations and possible modifications, which may improve sensor performance or the variety of operation.   
• Line 345: Report here the maximum flight altitude, the drone and sensor can achieve.
• Line 351: 100 Hz is not used for the atmospheric measurements. Change to 10 Hz.
• Line 353: “resolution” is used more for time or spatial resolution. Not for precisions! State other possible error uncertainties.
• Line 357: Was 6.25% reported in the results part somewhere?
• Lines 358-359: Yes, flux estimates closer to the ground had a ~50% contribution from this one field experiment, but this was derived from a model estimate. The drone did not fly that low. Please specify, and this statement is quite general for relying only on one example. Additional literature would help here.
• Lines 360-362: Great! But add those points already in the discussions in the previous sections, so that you can conclude those here but not state them for the first time.

Literature:

Lan et al., 2026, https://doi.org/10.15138/P8XG-AA10

Curtius et al., 2024, https://doi.org/10.1038/s41586-024-08192-4

Karion et al., 2015, https://doi.org/10.1021/acs.est.5b00217

Conley et al., 2017, https://doi.org/10.5194/amt-10-3345-2017

Riese et al., 2025, https://doi.org/10.1175/BAMS-D-24-0232.1

Müller et al., 2015, https://doi.org/10.1002/2014GL062556

Ort et al., 2024, https://doi.org/10.5194/amt-17-3553-2024

D’Amato et al., 2025, https://doi.org/10.1364/OE.558437

Viciani et al., 2018, https://doi.org/10.3390/s18072380

Röder et al., 2024, https://doi.org/10.1007/s00340-024-08254-5