General comments:

This manuscript presents comprehensive analysis of the use of different UAV systems and calculation methods to quantify methane points emissions in an Arctic seep area. The two UAV systems were equipped with open-path and close-path methane instruments, along with two different onboard 2D anemometers. This study further compares methane emissions calculated using mass balance approach and Gaussian plume inversion, finding that the mass balance approach provides more robust quantification with smaller uncertainties.

Overall, this manuscript is suitable for AMT. My main concerns relate to how wind measurements were compared between onboard and ground-based anemometers. Since different wind measurements were used to calculate emissions, it is unclear how the results can be directly compared when different anemometers were applied. Please refer to the specific comments below. I would recommend publication after consideration of the following comments.

Specific comments:

1. Line 100-110, The two UAV setups used different 2D anemometers. The onboard instruments measure relative wind speed measurements (apparent wind). Was true airspeed measured on the UAVs? What is the impact of using different 2D anemometers on the emission calculations? How were wind measurements compared between onboard and ground-based measurements? (Line 110). Please clarify.
2. Line 218, how was the background concentration determined? Could the background level vary with altitude? The background concentration appears to differ is differ between UAV-MPI and UAV-NRCan. How large is the uncertainty associated with background concentration when calculating emissions using different methods?
3. In addition to the temperature (Appendix A), how stable was the pressure during flight? How might pressure variations affect the measurements?
4. Was the impact of water vapor on methane measurements considered? Were any drying systems or correction equations applied?

Technical comments:

1. Line 147, please correct “whichis”

General comments:

This study deploys two UAV platforms equipped with open-path and closed-path methane analyzers to quantify emissions from a known geological CH₄ seep at the Mackenzie River Delta, Canada. Vertical curtain flights at two downwind distances are used to compare three flux quantification methods: direct mass balance, cluster Kriging mass balance, and Gaussian plume inversion, while evaluating the influence of sensor types on flux estimates. The paper reports a mean seep emission rate of 11.4 ± 6.8 kg CH₄ / hr across methods and concludes that direct mass balance yields the most consistent estimates across platforms.

The paper addresses an important problem: quantifying geological CH₄ seeps in remote Arctic environments where conventional monitoring is impractical. The study site is interesting, and the simultaneous comparison of three quantification methods to the same real-world dataset is a valuable contribution, as is the assessment of how open-path and closed-path sensor characteristics influence flux estimates. The manuscript is well-written and transparent about the assumptions underlying each approach.

My main concern relates to the scope of the conclusions relative to the dataset. The study comprises four flights at one site under one set of atmospheric conditions, with no independent ground-truth emission rate available. This supports an informative method comparison and sensor comparison for this specific case, but I find that the evidence is insufficient to conclusively support broader claims about method robustness or generalizability beyond this specific test case. Moreover, the three methods operate within different uncertainty frameworks, making direct comparison of uncertainty magnitudes difficult in my opinion. Furthermore, the two platforms differed not only in sensor type but also in flight strategy, making it also difficult to attribute differences in flux estimates solely to sensor characteristics.

I recommend reframing the manuscript as a detailed case study. The authors apply established methods under conditions that are likely more challenging than most previous applications, which typically involved temporally stable (industrial) sources, homogeneous terrain, and sources above ground level. Explicitly situating the results within this context would enable the authors to discuss method performance, sensor behavior, and experimental design insights in this test case relative to prior studies. This would make the manuscript a valuable resource for the community, especially for the design of future UAV-based methane flux studies of natural seeps. Correspondingly, I suggest revising the title as well.

Specific comments:

P1, L16-17: 

"finding that mass-balance approaches yielded the most robust quantification with smaller uncertainties": From my point of view, this conclusion is not sufficiently supported given the limited dataset. The three methods rely on different uncertainty frameworks, and the mass balance approaches omit an important source of uncertainty: "the turbulent nature of atmospheric transport" (P8, L173-175).  

P1, L18:

"average estimated rate": I would clarify here that this average is taken across different methods applied to four drone flights, as the current wording could be interpreted (as I did at first) as an average over repeated experiments. 

P2, L37-39: 

"Accurately estimating the emission rates from point or localized sources is only possible if the locations are already known": This statement may overstate the limitation as there are UAV-based measurement approaches that have shown the ability to locate and quantify unknown point sources simultaneously. 

P4, L91:

From my understanding, Dallimore et al. identify two seeps close to each other at the Channel Seep 2 site. It would be helpful to specify which seep is investigated here. This would make it clear whether the second seep was downwind or upwind, and address any potential concerns about whether the other source could have influenced the measurements.   

P4, L92:

"high ebullition rate": This description would benefit from quantification or a reference to an observed rate. Please state explicitly whether the source is assumed to be time-invariant over the measurement period, and discuss how plausible this assumption is.

P4, L92:

Because the seep is located within a water channel, I wonder whether channel flow could have introduced advective transport of CH₄, influencing the plume. Could the authors discuss this? 

P4, L109-111 and Table 1:

"Two additional ground-based wind sensors were deployed to verify the UAV-based measurements": Please provide more details on the sensor type, measurement height, and location. Furthermore, the cruising speeds of the UAVs were relatively high (2-3 m/s, 5 m/s, and 7 m/s), potentially affecting the reliability of wind speed and direction measurements. The statement "UAV-based wind speed and direction measurements showed good agreement with the ground-based measurements (data not shown)" is important and interesting. I suggest adding the ground-based wind data and a discussion of the observed agreement in the manuscript. Please also report the distance between the wind sensor and the rotor plane or body of the drone. 

P7, Eq. 2 & P13, L270:

Please clarify how the background mixing ratio is determined and what is meant by "sampling the background concentration once". It would be helpful to add whether and how its uncertainty is included in the overall uncertainty quantification.

P7, L156-157:

"At the lowest level of the sampling plane, we used a logarithmic function to complete the vertical profile, assuming zero flux at ground level (Bonne et al., 2024).": This is not intuitive to me, considering the source is at ground level. The flux profiles in Fig. 6, as well as in Bonne et al., are concave (bowing towards the upper left for z(q(z)). In neutral conditions, the wind profile is typically convex (bowing towards the lower right for z(u(z)). Since q(z)=u(z)c(z), a concave flux profile would require concentrations to increase with height. Bonne et al. studied elevated sources; however, in the present study, the source is at ground level, and high concentrations near the surface are expected (e.g., by the Gaussian plume model), implying a convex rather than concave profile is fitting. 

P7, L166:

"We quantified the instrument errors using error propagation based on field measurement data as well as laboratory tests". Please specify which components were included (e.g., instrument precision, wind speed uncertainty, background concentration uncertainty), and what values were assigned to each.

P8, L166-170:

"In the CKMB method, the uncertainties are quantified using the covariance matrices provided by the Kriging algorithms, which were below 10% of the calculated emission rates for all cases. For the DMB method, the uncertainties associated with linear interpolation along the vertical axis are estimated to be around 10% based on the Kriging algorithm uncertainties.": I wonder about the reasoning behind applying Kriging‑based uncertainty estimate to the DMB approach, which does not rely on Kriging. Can the authors please clarify this.

P8, L182:

"incorporating variable wind direction into the model": To my knowledge, temporal variability is not included in this model. Rather, it allows for fixed misalignment between the mean wind vector and the coordinate framework and computes dispersion differently from the Gaussian plume formulation. 

P8, Eq. 4:

Please clarify if the surface is treated as a flat, homogeneous surface. Dallimore et al. mention that "prolonged activity has caused an erosional niche to form on the river bank and the formation of a 5 m deep pockmark" at the Channel Seep 2 site. I wonder whether and how this topography could have influenced plume dispersion, and therefore affected the flux estimates. It would be helpful if the authors could discuss this.

P9, L225:

"The measured peak CH₄ enhancements for both OP curtains are 2-3 times larger than those measured in the CP curtains": I think that this is an interesting observation. The authors note in the introduction that OP analyzers allow "near-instantaneous response" but do not characterize the effective response time, smoothing, or time lag of CP analyzers. Is it possible to include anything about this based on the experiment? For example, Morales et al. characterizes the smoothing and lagging of the AirCore system (but both instruments were deployed on the same drone in their study). I wonder if prior studies also observed 2-3 times larger enhancements?

P11: 

I find the observation of asymmetric tails interesting. The authors attribute the asymmetry to two potential factors. By comparing data from transects flown in opposite directions, as shown in Fig. 5, the wind incidence effect can be isolated from the instrumental effect. The data appear to suggest that the instrumental effect is dominant. For future studies, I wonder whether the asymmetry can introduce a systematic bias in the flux estimates or whether flying alternating transect directions (partly) mitigates this effect. I think that a comparison with observations in the literature (if any, Morales et al.?) would strengthen the discussion. 

P13, L260-272 & P17, L314-320 & P17, L322-323:

"The calculated emission rates [...] agree within their estimated uncertainties.": I would avoid directly comparing uncertainty ranges across methods, as their underlying frameworks are fundamentally different. Furthermore, with uncertainty ranges of 45-54% for the mass balance methods and up to 89% for GPI, I think that these ranges are too wide to discriminate between methods based on a single test case and should not be used as evidence for relative method performance. Instead, I think that it would be more interesting to compare the observed method and sensor differences with those reported in prior literature.

Section 3, especially Section 3.3:

I suggest reframing the findings as insights into method behavior rather than as evidence of method superiority. Please consider adding a specific 'lessons learned' and/or 'outlook' section, as this could provide a lot of value to the community when designing future studies. 

P15:

Whereas curtain flights make sense for the mass balance methods, the flight path may not be the most informative for constraining the plume shape parameters in the GPI method. Other flight patterns may provide more information about the 3D structure of the plume. Consider adding this to the discussion.

P18:

I think that Dallimore et al. report an observed maximum gas seepage rate at the Channel Seep 2 site. Consider including their findings in the figure or text.

P19, L367-368: 

"If UAV-based wind measurements are available, mass balance approaches (CKMB and DMB) to quantification should be preferred over the GPI because they reduce the number of assumptions involved in the calculations.": To my knowledge, both methods share many of the same assumptions, importantly: spatial and temporal stationarity of the plume during sampling. MB methods require interpolation and extrapolation of the concentration and wind field, while GPI assumes a Gaussian profile. To me, it is not clear that MB methods are based on fewer assumptions than GPI. Moreover, I consider the number of assumptions to be less important than the total uncertainty and bias they introduce. 

Appendix A:

I think that this is an interesting finding, and the authors could consider moving it to the main body of the paper. 

Technical corrections:

P7, L147:

Typo "whichis". 

P20, L392:

Missing units.