Overall Evaluation

This study presents the first comprehensive assessment of the novel AVIRIS-4 imaging spectrometer for detecting and quantifying methane point source emissions. Through a blind controlled release experiment, the authors systematically evaluate the sensor's detection limits, emission estimation methods, and associated uncertainties. The research is rigorously designed, methods are thoroughly described, and the analysis is insightful. The observation and correction of the "plume shadow" phenomenon at high spatial resolution is particularly innovative. This work provides strong evidence of AVIRIS-4's capability for low-level emission detection and represents a significant contribution to the greenhouse gas remote sensing community. I recommend acceptance after minor revisions addressing the points below.

Major Strengths

(1) Comprehensive Methodology: The descriptions of the AVIRIS-4 sensor specifications, data processing pipeline, methane retrieval (MF/LMF), emission estimation (IME/CSF), and uncertainty analysis are systematic and facilitate reproducibility.

(2) Novel Findings: First observation and explanation of the "plume shadow" effect in high-resolution imagery, accompanied by a proposed correction method. Demonstration of a novel method to estimate emission source height from shadows.

(3) Practical Relevance: Clearly defines the sensor's detection capability down to 1.45 kg CH₄ h⁻¹ under ideal, low-altitude flight conditions, providing valuable guidance for practical monitoring applications.

(4) Robust Uncertainty Analysis: Provides a systematic evaluation of how wind speed, spatial resolution, and shadows impact emission estimates, correctly identifying wind speed as the dominant source of uncertainty.

Specific Comments

1. On Retrieval Algorithm Limitations

(1) Linearity Assumption: The matched filter (MF) and lognormal matched filter (LMF) are based on a linearized Beer-Lambert law, which may be invalid for the strong, localized enhancements from intense sources resolved by AVIRIS-4. The iterative approach refines the input spectrum but does not overcome this fundamental theoretical constraint. The authors should assess the severity of potential non-linear absorption effects for the strongest plumes (e.g., 290 kg h⁻¹) to better define the operational limits of their retrievals.

(2) LMF Background Bias: The observed background bias introduced by the LMF is a significant concern, likely stemming from the log-transform amplifying noise in low-SNR pixels. This is a fundamental signal-processing issue, not merely a challenge for automation. A comparative analysis of the noise covariance matrix (Ŝ) in linear versus log space for different surface albedos is needed to elucidate the mechanistic origin of this bias.

2. On Radiative Transfer at High Spatial Resolution

(1) Plume Shadow & Surface Heterogeneity: The plume shadow is a noteworthy finding. However, the geometric correction (Eqs. 6-7) assumes a uniform surface albedo. In real-world scenarios, the plume and its shadow often overlie heterogeneous surfaces (e.g., vegetation vs. asphalt), meaning the two light paths experience different ground reflectances. This will introduce errors. The authors should discuss this limitation and its potential impact on quantification accuracy.

(2) 3D Effects: At AVIRIS-4's sub-meter resolution, 3D radiative transfer effects, such as adjacency contamination from scattered light, may become non-negligible. The albedo artifact mentioned in Section 3.5.1 indicates such effects. The authors should comment on whether adjacency effects could influence retrieved methane enhancements, particularly within the core of strong, compact plumes.

3. Methods and Data Processing

(1) MF/LMF Comparison: The conclusion regarding LMF's performance should be supported by a systematic, quantitative comparison against the standard MF. Statistical metrics (e.g., RMSE, bias distribution across the entire dataset) are needed, rather than reliance on selective examples.

(2) Convergence Criteria: The convergence criteria for the iterative absorption spectrum calculation should be explicitly defined. Please specify the quantitative threshold (e.g., change in mean plume enhancement between iterations) used, rather than stating it converged in "2-3 iterations."

(3) Retrieval Justification: A brief justification for relying solely on matched filter techniques would strengthen the manuscript. Please comment on why more robust methods like WFM-DOAS were not considered, especially given the potential for non-linear effects from strong sources.

4. Results and Discussion

(1) Wind Speed Comparison: To complement Figures 7 and 8, a summary table with performance metrics (RMSE, MBE) for each wind speed method, stratified by emission rate bins, would provide a clearer and more systematic comparison.

(2) Quantifying Plume Shadow Impact: The discussion on plume shadows would be strengthened by a quantitative analysis. Providing specific values for the emission rate bias (comparing corrected vs. uncorrected estimates for affected plumes) is recommended.

(3) Resolution Trade-off: The trade-off made by AVIRIS-4 (higher spatial resolution at the cost of spectral resolution) warrants brief discussion. What are the implications for retrieving other gases (e.g., CO₂) or for applications over complex surfaces?

(4) LMF for Strong Sources: The finding that LMF offers no improvement for strong sources seems to contradict Schaum (2021), who posits it as the uniformly most powerful detector. This discrepancy should be discussed.

5. Uncertainty Analysis

(1) Error Typology: The decomposition of wind speed uncertainty should more clearly distinguish between systematic (e.g., σ_eff, σ_inst) and random (σ_var) error components, as their impacts on the final emission estimate differ.

(2) Pixel Area Uncertainty: The 5% uncertainty estimate for pixel area, based on visual comparison, appears subjective. A more objective quantification, potentially from an analysis of geolocation residuals using ground control points, is recommended.

6. Minor Points

Detection Limit Context: The abstract and/or conclusions should present a clearer, more direct statement comparing the detection limit of AVIRIS-4 with its predecessor, AVIRIS-NG, as the current phrasing is somewhat vague.

The manuscript by Meier et al. deals with a methane controlled-released experiment using the AVIRIS-4 airborne spectrometer. The authors report on the performance of AVIRIS-4 for the detection and quantification of methane plumes, and also use the generated dataset of plume detections for the illustration and evaluation of several technical aspects related to the remote sensing of methane point sources.

Overall I think it is a nice study with several important messages for the growing methane remote sensing community. The manuscript is well written and presented, and the topic fits perfectly in AMT, so I recommend its publication.

I would like to request the authors to address the following points in their revision of the manuscript:

- Plume shadows (section 2.4.4 and 3.5.4): the authors illustrate the issue of “plume shadows” in their dataset. However, the discussion of this effect and the implications for plume detection, attribution and quantification remain quite superficial. I would recommend the authors to deepen the discussion of this effect. In particular, the authors could discuss the preprint by Gorroño et al. https://egusphere.copernicus.org/preprints/2025/egusphere-2025-4924/, which is focused on this topic.

- AVIRIS-4-specific Ueff model (section 2.5 and 3.4): the authors test the Ueff model developed by Varon et al. for GHGSat. Even if they find that this model combined with the 10-m lidar wind speed helps reduce the offset between the reported and estimated Qs, I wonder if an AVIRIS-4-specific Ueff model should be used for this test? This could be one model trained for the specific conditions of these acquisitions, at least in terms of retrieval noise and range of emissions. Regarding the spatial resolution, I acknowledge it is not feasible to recreate the very high spatial resolution of the AVIRIS-4 observations in the training of the model, but even getting to a 25-m sampling might help further improve the results.

- MF vs LMF (sections 2.4.1, 2.4.2 and 3.5.5): the basic and log versions of the matched filter retrieval are compared. The authors find that “improved detectability can be attributed (...) to reduced noise levels in the CH4 retrievals,” (by the LMF), and that “the LMF had little to no effect on CH4 enhancements for the largest release events in the campaign”. This is actually interesting, as I would expect the opposite results: the LMF leading to a worse plume detectability due to the higher sensitivity to the surface and to it being more prone to generate false positives, but at least being helpful to correct the underestimation of XCH4 in the stronger emissions. Could you please comment?

- POD (section 2.7): I agree that this type of controlled release experiment including multiple overpasses can be very useful to establish POD functions for methane-sensitive instruments. However, I don’t understand the relationship of this section with the rest of the study. Is Eq. 21 a result from the analysis (hence it should be moved to Sec 3)? how is it related to the plume detection limits discussed in other parts of the study? Would it be possible to show the dependence of the detection limits on U10 based on this equation?

- Estimation of CH4 retrieval uncertainty: Eq. 13 is set to account for measurement errors propagated to retrieval errors, but wouldn’t it be better to estimate the 1-sigma retrieval error from the data themselves, e.g. as the StdDev of XCH4 over plume-free regions of the scene?

- Estimation of emission rates: both the IME and the CSF models are used, and one or the other are being selected depending on the plume morphology (“L300, We were able to estimate the emission from 67 of the 68 detected plumes, 54 of which were estimated using the CSF method and 13 using the IME method”). How is the method actually determined for each plume? Related to that, in L280 the authors write “the CSF method outperforms the IME approach, as the effect of turbulence is reduced through averaging across multiple cross-sections”. This is an important statement in my opinion, and it would be nice to see it more deeply discussed. Also, I think it would be very interesting to see a comparison of the Qs between the two methods for the plumes, both between each other and with the metered values.

- Estimation of emission rates without explicit use of external U10 data: one of the main conclusions of this study is that accurate wind speed information is key for accurate emission rate estimates (e.g. sections 4.2.1 and 4.2.2). I am wondering, does this call for the development of ML-based methods able to infer emission rates solely from 2D XCH4 plume maps, without specific use of external U10 information (e.g. https://www.sciencedirect.com/science/article/abs/pii/S0034425721005290, https://egusphere.copernicus.org/preprints/2025/egusphere-2025-1075/)? This could be discussed in these sections

Other minor comments:

- L1: what is a “critical greenhouse gas”?

- L2: I would replace “use” by “sector”

- L34: CO2M is not yet flying, unlike the rest of the missions mentioned in this paragraph.

- L101: “Observations over dark surfaces (…) have a low SNR

- Fig. 11: could the pixel size of each flight altitude be added?

- L428 “We assume that this technique can be applied if the length of the shadow is at least twice as large as the uncertainty in the length of the shadow”. I don’t understand what the basis for this statement is.

- L461-465: I think this type of information about AVIRIS-4 should be included in the Introduction section

- L483: “the linearisation of the unit absorption spectrum no longer holds and assumed enhancements for the calculation of the absorption spectrum have greater influences on the retrieved enhancements” – has this been shown in this study?