Rev of ACP-2021-857

Jens Muhle

This manuscript reports a decade of measurements of perfluorocyclobutane (c-C4F8) from the remote AGAGE sites.  c-C4F8 is a long-lived potentially important greenhouse gas.  The measurements accurately define emissions, indicating they are clearly northern hemisphere and correlated with the production of HCFC-22.  The analysis shows that c-C4F8 is a likely by-product and from the observations they calculate an emission factor.  The authors argue that better process management could reduce these fugitive emissions, which are much larger than the direct intentional production reported to the UNFCCC.  This work is an important extension to the Mühle et al. (2019) work on this gas.

The manuscript is well written (except see below) and provides a valuable contribution to our understanding of the synthetic greenhouse gases and their potential role in climate change.  The authors need to focus more on what is new here and not confuse with A5, non-A5, and China (3 different entities?).  Further, the data archive must be upgraded before publication.

L30 – Glad to see a clear chemical definition, noting the range of names used to describe c-C4F8.

L84 – Can you comment on whether the pollution events that were removed contained high levels of c-C4F8? Thus indicating nearby production?  Which stations?  This might be useful.

L120 – It would be good to include the black diamonds in Fig 2 in the legend, making it clear that they are the global c-C4F8 emissions. BTW, is it clear that China is neither A5 nor non-A5?  Is it being double counted here and in the Table?  This is confusing.

L135 – I do not understand the effort at linear fitting in Fig. 3.  It is confusing.  The TOTAL (green) fit makes some sense, presumably implying a source of ~0.14 Gg/y that is from non-HCFC-22 sources, and your yield of 0.31% (kg/kg).  The blue fit of global to only A5 emissions does not make sense.   The explanation in L126-155 does not help.   Why do you come up with an emission factor assuming that only A5 countries have processes that emit c-C4F8.  If this is the case, then why fit to global?  Please introduce the logic of this approach if you think it relevant.

L174ff – Again, this is confusing with the separation of China and comparing it the A5 and non-A5.  Discussing China, and then eastern China is more confusing as to what is measured and what is known.

L181-201 – This conclusion section is again very confusing for this reviewer.  It mixes discussion of the processes in great detail with the regions.  It brings in A5 and non-A5 and then discussions Eastern China, China, and Russia, without any understanding as to why these are important.  I think the authors want to emphasize the Muhle et al (2019) regional attribution with this global one.  I would de-emphasize the previously published work and focus on the A5 and non-A5.  Then separately discuss China if you want but do not confuse with the A5 and non-A5 modeling here.

L202-206 – This is a clear summary, the above is confusing.

*L210 – I am afraid that the data availability is really inadequate and so ‘old’.  You must post on a DOI the datasets used here, particularly Table 1 and Figures 2 and 3 so that others can compare readily with this work.  While AGAGE seems to live forever, it is not appropriate to list you data from an non-ODI source like your own website.  Please use a regular doi, I am that MIT must have one.

The manuscript by Mühle et al. investigates the relationship between calculated global emissions of PFC-318 and estimates of HCFC-22 production used as feedstock for production of fluorinated ethylene products used to manufacture Teflon.  This manuscript extends previous work (Mühle et al., ACP, 2019) by updating several years of PFC-318 emission and by looking in more detail at the use of HCFC-22 as a feedstock.  The material is well-presented, the underlying data on PFC-318 temporal variation is excellent, and the model used to evaluate emissions has been applied successfully to other AGAGE data.  I am not sufficiently expert to evaluate the accuracy of the HCFC-22 feedstock production data, which is the other significant compilation of this manuscript.  Any errors in these reported production numbers will impact the conclusion, and, if relevant, the authors should comment on the reliability of these reports (especially since inconsistencies are noted for other gases regulated by the Montreal Protocol).  

The major contribution of this paper is the observed linear relationship between HCFC-22 feedstock production and PFC-318 emission rates after about 2002.  Attribution of the increasing PFC-318 emission due to industrial activity in China is also interesting.  However, I thought the manuscript could have done a better job in describing why the relationship observed after about 2002 or so, is different from the preceding 10 years, and what was contributing to PFC-318 emissions prior to 1990.   The unknown nature of the period prior to 2002 adds a level of uncertainty to the analysis that is not discussed.  Between 1990 and 2002, the calculated PFC-318 emission rate remained nearly constant (even decreasing early in the period) while cumulative HCFC-22 FS production totaled about 1500 Gg.   For approximately the same HCFC-22 FS production from 2005-2010 (1405 Gg), the calculated PFC emission rate increased from 0.96 to 1.43 Gg/yr.  Similarly, the last part of the record shows constant HCFC-22 FS production but continued increase in PFC-318 emission.  How is this rationalized?  I would like to understand the different emission response of PFC-318 to HCFC-22 production to have better confidence in the quantitative results presented. The correlation presented looks compelling but deserves more detailed analysis of potential factors that might contribute to the correlation.

It might have also been interesting, if the data is available, to discuss the relative importance of HCFC-22 FS production to other sources of HCFC-22 and to the global burden and trend of HCFC-22.  AGAGE data shows a consistent linear increase in PFC-318, while the rate of HCFC-22 increase has slowed. Does this mean that HCFC-22 FS emissions are becoming a more significant component of ambient HCFC-22 relative to other changing sources?  In a similar vein, the authors mention CHF3 associated with the production of HCFC-22, and thus it has a source related to PFC-318 emission.  Can temporal trends in CHF3 vs PFC-318 observed at AGAGE sites provide additional insight into the relative importance of the different source emissions over time?  Data from at least one of the AGAGE sites (THD) seems to show a change in slope of the relationship between CHF3 and PFC-318 occurring around 2015.  It seems a shame not to get better use of the excellent measurements from the AGAGE sites.  This may well be beyond the scope of the manuscript intended by the authors, but it could make the manuscript more interesting.

Other minor comments/questions:

Line 84: Given the locations of the sites used for the 12 box model, it brings up the question of how much impact was observed in the pollution events that are removed from the data set for further analysis?  And, if significant, could further analysis of these pollution events provide further insight into regional or long-range sources?  Though not critical, it would be interesting to know the fraction of data that were removed due to pollution filtering.

Line 110: The authors note that the results here agree with Mühle et al., 2019.  Aren’t the data, model, and methods identical between this study and earlier?  I thought only the last few years of data were new.

Figure 2: The black diamond curve is not identified in the caption or figure as the PFC-318 emission rate.

General Comments:

This paper represents an important advance in our understanding of the sources of global perfluorocyclobutane emissions. These emissions matter. As discussed in this paper and in Mühle et al (2019), perfluorocyclobutane is a long-lived, potent synthetic GHG whose emissions are rapidly growing. Other data sources also indicate that perfluorocyclobutane emissions are significant. Data from the US Inventory of GHG Emissions and Sinks and from the USEPA Greenhouse Gas Reporting Program indicate that perfluorocyclobutane is the third most emitted perfluorocarbon in the US in GWP-weighted terms, behind perfluoromethane and perfluoroethane. Perfluorocyclobutane is also consistently among the top two or three most emitted fluorinated GHGs of any kind from US fluorochemical production in GWP-weighted terms.

The data and analysis presented in the paper make a compelling case that pyrolysis of HCFC-22 is the chief source of perfluorobutane emissions globally. This case begins with the detailed and richly sourced discussion of the pyrolysis process and the role of perfluorocyclobutane in it and proceeds through the author’s calculations of global emissions of perfluorocyclobutane and discussion of HCFC-22 feedstock production in developed and developing countries. The authors’ representation of the pyrolysis process and its generation of perfluorocyclobutane is consistent with discussions that this reviewer has had with US producers of HFP and TFE. Their analysis of UNEP and TEAP reports of HCFC-22 production for feedstock use, and the correlations between this production and perfluorocyclobutane emissions, is thorough and thoughtful. Their interpretation of these trends and relationships is persuasive but nuanced.  

Specific Comments:

To put the inferred global c-C4F8 emissions into perspective, it would be helpful to compare them briefly to inferred global emissions of other long-lived GHGs, such as SF6, CF4, and C2F6, both in unweighted and GWP-weighted terms.

On page 2, line 35, recommend deleting “regulated and” and “the Kyoto Protocol of,” resulting in the following sentence: “Emissions of c-C4F8 from developed countries are reported under the United Nations Framework 35 Convention on Climate Change (UNFCCC).” The requirement to report emissions of c-C4F8 is what is relevant here, and it applies to countries that are signatories to the UNFCCC but not the Kyoto Protocol.

On page 2, line 58, the authors mention electrochemical fluorination (ECF) as a potential alternative means for manufacturing TFE and HFP that would generate less waste than the currently dominant method. This was surprising, because at least some types of ECF are known to generate large quantities of waste products. This reviewer is not familiar with the research cited by the authors in support of their suggestion to explore ECF as an alternative (Ebnesajjad and Mierdel), which may be based on more recent and refined types of ECF. I recommend qualifying the recommendation, perhaps by inserting “refined methods of” before “electrochemical fluorination” on line 58 and in other places where ECF is mentioned.

Figure 2 on page 6 contains a wealth of interesting data that is generally thoroughly discussed in the paper. Particularly fascinating was the analysis of using the 1996 through 2000 feedstock and emissions data to estimate an emission factor for developed countries, where the authors concluded that the current emission factor from developed countries was probably lower than the average from that period. However, there was a sharp divergence between c-C4F8 emissions and global feedstock production in 2009 that the authors did not mention. This is sufficiently striking that it is probably worth discussing briefly, even if the conclusion is that the cause of the divergence is not known. (The same is true of a smaller divergence in 2016.)

On page 8, the authors state “Current industry knowledge is that less than 2% of HCFC-22 FS produced is used in reactions that do not involve the TFP/HFP/c-C4F8 route,” but they do not cite a source. A source should be cited for this statement because the correlation between HCFC-22 feedstock production and TFP/HFP/c-C4F8 production is fundamental to the method used by the authors to reach their conclusions.

Technical Corrections

Page 2, line 55: “TFE producer” is missing an “s.”

Figure 2: The legend is missing an entry for the black diamonds, which appear to represent global inferred c-C4F8 emissions. In addition, the black diamonds appear to be offset from the other icons in the chart, perhaps because the values shown are from the middle of the year. Whatever the reason for the offset, it should be explained.

Figure 3: The legend currently appears very close to the data, making it look as if the icons being explained in the legend are part of the data. Recommend either moving the legend or enclosing it in an outline so that the reader can distinguish between the legend and the data.

This paper provides a valuable addition to the literature linking the pyrolysis of HCFC-22 feedstocks to c-C4F8 by-product emissions measured in the atmosphere.  The paper is well-written.  The explicit comparison to global feedstock supplies and derivation of a global emission factor is new and support the conclusion that pyrolysis of HCFC-22 is the dominant source of c-C4f8 in the atmosphere.  The link to feedstock production in developing countries is interesting and provides a useful estimate of overall emission factor from the sector in developing countries.

Specific comments:

The paper would however benefit from discussing other sources of c-C4F8 in a quantitative manner if possible.  The intercept from Figure 3 from global production of HCFC-22 feedstock suggests a significant other source.  Emissions of c-C4F8 from semiconductor manufacturing, largely from consumption of c-C4F8 for etching processes, are small but not insignificant and could account for a large portion of the background.  In 2018, WSC reported emissions of ~0.13 Gg (https://www.eusemiconductors.eu/sites/default/files/23rdWSCJoint-Statement_May2019Xiamen-TOC_FINAL.pdf).  Total emissions of c-C4F8 from the electronics sector may be larger, as this estimate does not include emissions from PV, LDC or MEMS.  Use of c-C4F8 in the semiconductor industry started increasing at about the same time as the increase of C4F8 in the atmosphere (early 2000s – see Francesca Illuzzi & Harry Thewissen (2010) Perfluorocompounds emission reduction by the semiconductor industry, Journal of Integrative Environmental Sciences, 7:sup1, 201-210, DOI: 10.1080/19438151003621417 ).  On page 8 of this manuscript, you refer to the possibility of significant emissions from the semiconductor industry between 1996 and 2001, but this seems unlikely.  No emissions of C4F8 were reported to the US EPA prior to 2002.

Considering that there was a large dip in production of HCFC-22 in 2009 but an increase in emissions, this may be evidence that the by-product emissions from Annex 5 countries and China is greater than for developed countries, as these two regions had only a minor decrease or an increase (China) in 2009 or alternatively that there was not a corresponding drop in production for other c-C4F8 sources.  There was also a dip in production in 2015 (China + A5) but no corresponding dip in emissions.  Appears to instead be large increase.  There was a corresponding increase in c-C4F8 emissions in S. Korea in your previous paper (Muhle 2019) and in US emissions from the semiconductor industry (EPA 2020).  Even if the authors are not sure of the source of these differences in trends, it would be useful to discuss the possible sources. 

Other typographical/formatting comments:

For Figure 2, it would be useful to include the black diamonds in the legend.  It also appears that the emissions data is offset on the x-axis compared to the production data, making it appear that there are two additional years of emission data compared to production data (but there is only one additional data point).  It would be easier to read if they were not offset.