In their work Conrad and Johnson handle the importance of error correlation in wind speed data when deriving Methane emissions from measured concentrations. An algorithm to quantify spatiotemporal auto-correlation is described, giving guidelines on how to best perform measurement campaigns in certain regions of interest. While this study mainly focuses on the methane emission example, the core method is applicable for any method that relies on model wind data, further underscoring the scientific significance of this work. Overall, the presentation quality of this study is excellent. In the following some minor revisions and technical corrections are suggested that mainly focus on improving the understand-ability of the study.

Minor revisions:

• As a reader who is not proficient on measurement statistic algorithms, section 2 would largely benefit from a more tangible explanation approach using less mathematical detail and more graphics that explain the used methods. It would be very helpful to have a clear recipe of what is needed to apply the described algorithm (e.g. NWP data on a fine grid with high temporal resolution and statistically independent station data).
• While the authors give a clear explanation on the importance of error calculation for the resulting Methane emission estimate, an estimate on how the described method compares to other sources of uncertainty could provide more insight in said importance. Examples for other sources of uncertainty in emission calculation: Injection height and resulting usage of the wind field (speed and direction), missing “measured” wind data in the atmosphere above ground level, uncertainty of the measured concentrations.
• Section 3.5 nicely shows the effect of model resolution on the wind speed error model. However, the aforementioned comparison to other sources of uncertainties could provide information on how the importance of the described model changes for different spatial or temporal resolutions of the NWP data set. This would also help the reader to understand the importance of the wind speed error model.
• A methane emission calculation comparison between the following three approaches would further the understanding of the importance of using a wind error model: detailed handling of error calculation (main topic of this study), a simple approach to error handling (probably similar to RER approach described in the study) and the approach of neglecting wind error.
• The work motivates why a model of the wind speed error is important and how to best apply the gained knowledge, e.g. in planning of measurement campaigns. However, I’d like to see at least a small focus on how to handle imperfect conditions: What do I do if I don’t have an independent measurement data set in addition to a NWP using data assimilation? Is it possible to generalize some of the found features? Maybe using parameters like surface roughness, main wind direction and topography?

Technical corrections/suggestions:

• Page 1 Lines 20—24: Instead of providing the finding of how to best perform measurements w.r.t. correlation, the estimate on how large the emission uncertainty increases if neglecting wind speed error correlation would in my opinion be beneficial for this study.
• Page 2 Lines 3—24: I’m missing a step in between describing the common challenge and why/how much the correlation of “wind speeds (and hence their uncertainties)” affects the emission. Maybe the authors could give an example emission calculation from given methane enhancements. This could help to better explain where in that calculation, correlation of measurements and underestimation of the wind speed error affect the derived emission.
• Page 2 Line 11: I had difficulties finding the work from Branson et al., 2021. The other example for an aerial measurement approach (Thorpe et al., 2021) describes methane emission estimates using a LiDAR technique, while LiDAR is separately mentioned in the second half of the sentence. The currently sentence suggests that these two methods are different, but the references point to the same measurement technique.
• Page 17 Line 11: […] of their semivariogram (left axis) and the their correlogram […] - remove the "the" after "and"
• Page 18 Line 2: […] the spatial correlogram is trivially calculated by […] - remove the "trivially" after "is"
• Page 19 Lines 7/8: […] At large lags, temporal correlations, representing bias over the diurnal cycle, oscillate with an amplitude of approximately 0.13. […] – is there a physical reason for this diurnal bias? Is it connected to sub-model-scale meteorology?

In this manuscript, Conrad and Johnson discuss the importance of wind speed uncertainties when estimating methane emissions using remote sensing observations. The study is strongly focused on theory and technical elements of model development. They demonstrate the applicability of a combined model for wind uncertainties for a testcase in Canada. They also suggest a new experimental design to reduce temporal autocorrelation effects. The topic is of great importance as this source of uncertainty is usually neglected in remote sensing studies. The text is well written und the manuscript has a clear and easy to follow structure. However, given the focus of the manuscript on model design it might be better suited for another EGU journal, namely, Geoscientific Model Development.

Nevertheless, the study also overlaps with the scope of AMT.

There are only some minor comments that should be addressed before publication.

Page 1, line 27: Would you consider the level of uncertainty similar for aircraft and satellite studies or should they be considered differently? Especially the fact that airborne surveys often have on-board wind data and are not restricted to clear-sky day bias could suggest that they might not experience the same limitations.

Page 2, line 17: Why is the analysis limited to May to October? Satellites are observing and reporting observations in all seasons. Are you confident there is no seasonal bias in the NWP performance?

Page 7, line 14-15: The two references cited here: Sklar 1959 and Nelsen 2006 are not easily accessible or behind a paywall. So, please provide more details on Copulas here or provide additional references that discuss Copulas in more detail.

Page 9, line 12: More details on DECLUS would be helpful here.

Page 9 line 19: If there is a myriad of literature, why do you only provide a single reference, which is, again, behind a paywall.

Page 12: Section 2.3. highlights that this study is really about the model itself and maybe better suited for Geoscientific Model Development. Nevertheless, it is a good example for more detailed analysis of correlated uncertainties affecting many applications.

Page 19, line 24: The point about sun-synchronous satellites is crucial and it might be good to highlight that nearly all current satellites used for methane emission monitoring are sun-synchronous.

This manuscript provides a rigorous probabilistic modeling study of numerical weather prediction (NWP) 10-m winds' error relative to independent ground-based wind speed measurements. A region-average wind error model is fitted as a Weibull distribution. Then, the errors are modeled marginally in space and time and combined using Gaussian coupla and spatiotemoral semivariograms. It seems technically sound and well written, with the following points to consider for further improvements.

A main issue is that the connection between the core analysis/results of this work and the scope of AMT (specifically remote sensing of methane point source emissions) is relatively weak. It is really section 3.5 only. It may be helpful to provide more context on how this work fits in the methane quantification pipeline. Specifically, the 10-m wind that this study tries to model, as the ground truth, is measurement at (usually) 10 min interval. The 10-m wind->effective wind->methane emission pipeline is calibrated all in an LES model in Varon 2018. To what extent is the "ground-truth" wind speed relevant, if the wind-emission relationship is calibrated by a model? This study tackles the error from NWP to ground-measured 10 m wind, and it would be nice to have some discussion on the mapping from measured wind to effective wind and then to emissions. It almost makes me feel that targeting measured wind is a detour, and one should map NWP wind to LES wind, or whatever wind that calibrates the wind-emission relationship.

Fittings of the regional wind error model and the semivariograms are important for this work, so it is recommended to include details on how those fitting algorithms, specially the numerous constraints in fitting parameters, are implemented.

Technical comments:

Eq. 8: please double check as π seems to be reserved for a PDF, and dividing a PDF by u~ is unlikely to give another PDF.

Page 9, lines 10-11: it is recommended to provide the formula of AIC as the objective function.

Page 10, line 1: should that be the "... the variance of the difference of the zi at these positions"?

Page 11, lines 9-10: double check the location and span of bins. Should that be 0-50 km and 25-75 km?

Page 11, lines 24-25: H seems to be reserved for joint CDF. Consider another font/symbol for Heavyside step function.

Page 11, lines 26: please confirm what γ_k⁰(b;b) is, and why it should be 0.95.

Page 12, lines 1-2: fixing b3 to 0 contradicts the constraint (b3>b2>b1>=0) in the previous page. Please elaborate.

Page 12, lines 17-18: close the parenthesis.

Page 21, line 5: should the RMS uncertainty be read as the "distance" of the upper and lower error bars?

Section 4.1: it reads a bit strange to have such a dominant subsection in conclusion section. Consider making it a dedicated section preceding the conclusion.