Review of the article: “Continental-scale contributions to the global CFC-11 emission increase between 2012 and 2017” by Hu et al.

The manuscripts presents inverse modelling results for emissions of CFC-11 at a continental scale in order to better determine the source of a prior-reported global emissions increase. It uses measurements from two aircraft campaigns (for two different periods) and measurements from NOAA’s high frequency and flask network under an emsemble of measurement and a priori emission scenarios. This manuscript will be well received due to the mystery of the source of increased CFC-11 emissions, likely due to unreported production. My only major concern is that the prior uncertainties are not clearly communicated, and as such it is impossible to deduce whether the inversion is actually providing any new information. This must be seen before conclusions can be judged. I hope this is just a matter of clarification, and I hope to see the eventual publication of the manuscript.

Below are a number of suggestions for revisions to improve the manuscript, followed by technical comments, before the paper is suitable for publication.

1. Section 2.2.3: This section can be cut down drastically. Rather than providing information on all approaches tried, simply state the one approach taken in the work. It is useful to hear the alternatives used, but is much better suited to relegation to supporting information.
2. Section 2.2.4: The use of e.g. “population_87_NA” is confusing as the reader is provided with no information of what that is. They also do not appear anywhere else in the text (perhaps save the definitions in an expansion of the caption for Fig 2). This whole section would be much clearer without the use of the “population_XX” terminology. Further, “prior emissions” are a distribution, not a single value. What you are referring to here is the prior mean(/median/mode as it’s Gaussian). ‘A priori’ may be another useful term if defined. This applied throughout the manuscript. There is also no information on the uncertainty given in your prior distribution, which makes it impossible to understand the information content/uncertainty reduction provided by the measurements. I suggest providing this information in Table 1 and Fig 6/7.
3. Lines 366-375: It is stated that r² is less than 0.1. It’s unclear which definition of r² is used here: there are multiple, some of which can yield negative r². As such, it’s not clear whether there is in fact negative correlation between the regions.
4. It is not fully clear to me how the emissions and uncertainties are derived from the ensembles. In Table 1, I assume that the ‘range of the best estimate’ refers to the range in the MAP solution across the ensemble members? I don’t understand the definition of the 2^nd set of uncertainties – is this the range of the lower/upper 2-sigma values? This needs a clearer explanation in the text. It’s not currently clear to me whether the posterior uncertainty for a given ensemble member is greater or smaller than the ranges presented.

Technical comments:

Abstract: ACP guidelines state that there should be no references in the abstract unless urgently required. I do not believe that to be the case here. Acronyms defined in the abstract should be defined again in the main text.

Line 17: “early detection” is subjective. Many wouldn’t consider 8 years later as early detection. Please amend.

Line 19: “parties to the MP”

Line 60: “policy makers and industrial experts”

Line 66: What is the definition of ‘eastern mainland China’?

Line 79: NOAA should be defined

Line 81: What’s the definition of “regional-scale” here?

Line 86: An outline of the structure of the paper would be useful here.

Line 91: Were the exact authors included in all studies defined as ‘ours’? If not, delete ‘our’.

Line 93: Upwind from the measurement location

Line 94: Change “is footprints or” to “, termed footprints, are”.

Line 96: State that these models are HYSPLIT and WRF-STILT.

Line 97: Pedantic, but Bayesian inversions always require prior probabilities, regardless of the constraints.

Line 104: Here z is an enhancement. Please reference here the later section on how you derive the enhancement from the measurement?.

Line 109: Again pedantic, but “maximum likelihood estimation” is by definition non-Bayesian. If Bayesian, you are finding the maximum a posteriori (MAP) solution.  

Line 130: Be consistent with the use of hyphenation of in situ/in-situ (I would remove it).

Line 155: Change ‘our’ to ‘the’

Line 156: relative to what? Best to give an approximate distance for how [far] away the sampling is from emissions.

Line 219: How long were they run back in time?

Line 344: Change “described above” to “described in Sect. 2”.

Line 363: It seems incorrect to me that 16 ± 10 Gg/yr then only has a percentage uncertainty of 80-90% of the global total.

Line 411: Change “given we’ve already know[n]” to “given it is known”.

Line 431: How are you defining the ‘best estimates’?

Manuscript ID acp-2021-793

Title: Continental-scale contributions to the global CFC-11 emission increase between 2012 and 2017 (Hu et al.)

This study analyzed the atmospheric CFC-11 measurements from two global aircraft surveys - the HIPPO (2009-2011) and Atom (2016-2018) campaigns, to estimate regional scale CFC-11 emissions and the emission changes between two campaign periods. The manuscript demonstrates how a well-designed aircraft measurement can be used to constrain regional emissions estimates.

Overall, the writing and figures are clear, and the methodology maximizes the functionality of high-quality datasets. I encourage the publication of this important work, with only a few minor considerations suggested below.

General comments

1. Global emissions: authors provided their estimates of global CFC-11 emissions for 2009-2011 and 2016-2018 periods in table 1. They were very briefly mentioned in lines 351-355. Authors need to describe more in details how they were determined, which datasets were used for the analysis, and how well consistent they were with the estimates from other studies.
2. Data selection: authors stated that the data included most of the aircraft profiling sampling below 8km. Then does it mean that the HYSPLIT model used to simulate footprint for inversion was also run from the surface boundary layer up to the 8-km altitude? How were the uncertainties associated with the HYSPLIT model analysis analyzed?
3. Prior emissions: since CFC-11 is an anthropogenic compound, it is reasonable to take population density-based distributions of the global CFC-11 emissions of 67 Gg/yr as prior emissions. But as a base case, authors may need to consider including area-based distributions.
4. 6-8: it was stated that the error bars for the emissions changes between the HIPPO and ATom periods were calculated from the sum of 2𝜎 errors derived for the HIPPO and ATom inversions. But propagated errors from a subtraction can be determined by the square root of the sum of the squares of each error. So, the errors shown in the lowest panels of Figs. 6-8 might be overestimated.