Please refer to the attached file.

Krol et al. presented a comprehensive analysis of interpreting NO₂ plumes over several power plants across the globe. High-scale turbulent mixing is realized by LES with a condensed chemistry scheme representing key reactions associated with NOx chemistry. What is particularly neat about this study is the diagnosis of the relative importance of atmospheric transport/mixing versus chemistry (via the covariance between NO₂ and OH) and their interaction as plumes disperse. They also examined how corrections of AMF and upper columns affect the model-data comparison. Overall, the manuscript is informative and well-organized. I enjoyed reading it and thereby recommend publishing it after addressing a few comments. 

General Comments:

• Title: The title of the manuscript centered on the estimate of NOx emissions, but the current manuscript seems to focus less on the inversion/optimization of NOx emissions, and more on the forward simulations of NO₂ columns and how transport, chemistry, and adaptation to TROPOMI affect the simulations. I may suggest the authors revise the title to better highlight the real novelty of their study.
• Methodology: The definition of plume background area and the choice of prior emissions seem to be unclear. Did the authors use “correct” emissions reported from the CoCO2 project? If not, since emission, mixing, and chemistry can all be uncertain, it would be hard to evaluate the model performance (especially given the end goal of estimating emissions). Also, the “background” or background air has been mentioned but not properly defined. Was there a fixed threshold or a fixed distance from the stack? The background definition may affect the statements about lifetime diagnoses (page 18).
• Results: I understand the authors’ great effort in adding chemical reactions in LES for several simulations, but the explicit estimation of uncertainties associated with transport/mixing and chemistry seem to be missing from the manuscript, especially when the authors also acknowledged large uncertainties in the chemical model parameterization and atmospheric transport (L43 – L45). Again, if the final goal is to estimate/optimize emissions as suggested by the title, it would be important to acknowledge uncertainties in every model component (using either a simple CSF method or more complex chemical inversions). For example, Wu et al. (2023) (as mentioned on L44) tested the WRF-Chem-based chemical parameterization by estimating the chemical uncertainties with NOx lifetime and NO₂-to-NOx ratio and by evaluating against TROPOMI using EPA-reported power plant emissions.

Minor comments:

Eq. 2: What are some other processes? Emission and deposition?

There seemed to be some considerations of the injection height (L78) and the use of a plume rise calculation (Table 5). But whether/how both injection height and the plume-rise were implemented in the model is not super clear to me, despite the acknowledgment and discussions in Sect. 4.4. More clarifications would help the readers, as injection height and plume height are important for point sources with intensive emission rates.

Future application: Just out of general curiosity, how is the possibility of applying this approach to urban areas?