Global increase of ozone-depleting chlorofluorocarbons from 2010 to 2020

The production of chlorofluorocarbons (CFCs) that would ultimately be released to the atmosphere was banned globally in 2010 under the Montreal Protocol. Here we use measurements combined with an atmospheric transport model to show how atmospheric abundances and emissions of five CFCs increased between 2010 and 2020, contrary to the goals of the phase-out. The Montreal Protocol allows CFC production for use as a feedstock to produce other chemicals. Emissions of CFC-113a, CFC-114a and CFC-115 probably arise during the production of hydrofluorocarbons, which have replaced CFCs for many applications. The drivers behind increasing emissions of CFC-13 and CFC-112a are more uncertain. The combined emissions of CFC-13, CFC-112a, CFC-113a, CFC-114a and CFC-115 increased from 1.6 ± 0.2 to 4.2 ± 0.4 ODP-Gg yr -1 (CFC-11-equivalent ozone-depleting potential) between 2010 and 2020. The anticipated impact of these emissions on stratospheric ozone recovery is small. However, ongoing emissions of the five CFCs of focus may negate some of the benefits gained under the Montreal Protocol if they continue to rise. In addition, the climate impact of the emissions of these CFCs needs to be considered, as their 2020 emissions are equivalent to 47 ± 5 TgCO 2 .


Increasing global abundances and emissions
Atmospheric mole fractions of CFC-13, CFC-112a, CFC-113a, CFC-114a and CFC-115 have been previously reported 8,9,17,18 and are updated here. Abundances of CFC-13 and CFC-113a had been increasing throughout their published measurement record (1950s to 2017), while those of to the disposal of appliances with foam containing HCFC-141b. Increasing or persistent emissions of several other CFCs 7-9 and HCFCs 10,11 have been attributed, at least in part, to their production as feedstock or as an intermediate/by-product during production of other compounds, such as hydrofluorocarbons (HFCs). Feedstocks are the raw materials used to produce another compound and are often assumed to be near-fully converted into the target compound. While the Montreal Protocol does not ban the use of chemicals as feedstocks or emission of some by-products, it does encourage best efforts to reduce emissions.
In this Article, we focus on five CFCs with few, or no, known current uses: CFC-13 (CClF 3 ), CFC-112a (CClF 2 CCl 3 ), CFC-113a (CF 3 CCl 3 ), CFC-114a (CF 3 CCl 2 F) and CFC-115 (CF 3 CClF 2 ), which have atmospheric lifetimes ranging from 52 to 640 yr 12,13 . We present measurement-based atmospheric abundance time series of the five CFCs, from which we derive global emissions using a model of global atmospheric transport. The atmospheric abundances and emissions of all five CFCs have increased since the phase-out for dispersive use of CFCs in 2010. The increase in emissions has resulted in a modest impact on stratospheric ozone  Records before 1978 are based on measurements of air extracted from polar firn rather than air collected in situ). Furthermore, atmospheric abundances of CFC-112a and CFC-113a were growing faster in 2020 than at any point in their atmospheric records. For CFC-13 and CFC-115, two independent measurement networks support the record-high mole fraction in 2020 (see Extended Data Fig. 1 and its caption for a more detailed discussion and Extended Data Fig. 2 for additional independent measurements of CFC-13 and CFC-115). For CFC-112a, CFC-113a and CFC-114a, it is worth noting that their isomers (CFC-112, CFC-113 and CFC-114), which are present in the atmosphere at larger abundances in all three cases, have continued to decrease in abundance throughout 2010-2020 (Extended Data Fig. 3).
Combined global emissions of CFC-13, CFC-112a, CFC-113a, CFC-114a and CFC-115 increased from 1.6 ± 0.2 to 4.2 ± 0.4 ODP-Gg yr −1 (ODP-Gg, mass weighted by their CFC-11-equivalent ozone-depleting potential (ODP)) between 2010 and 2020 (Fig. 2). The mean growth rate of these emissions is around 0.3 ODP-Gg yr −1 per year. Global emissions of CFC-11 increased between the periods 2008-2012 and 2014-2018 5,19 which were attributed to unreported production. The increase in global emissions between 2010 and 2020 of the five CFCs reported here (expressed as ODP-Gg yr −1 ) is around a fifth of the global increase in CFC-11. In terms of impact on climate, the five CFC emissions derived for 2020 are equivalent to 47 ± 5 TgCO 2 -equivalent (CO 2 e) yr −1 in 2020 (around 150% of London's CO 2 emissions in 2018 20 based on 100 yr global warming potentials). In 2020, combined emissions of CFC-112a, CFC-113a and CFC-114a

Unintentional CFC production
Previous work has linked emissions of CFC-113a and CFC-115 to the production of HFC-125, a compound used primarily in refrigeration and air conditioning in rapidly increasing quantities in recent years 8,9,17,21 . CFC-113a and CFC-115, along with CFC-114a 22 , are considered 'undesirable by-products' during HFC-125 production, which may account for at least some of the observed emissions. During HFC-125 production, up to ~1.1% by mass of the perchloroethylene (C 2 Cl 4 ) feedstock may end up as CFC-113a, ~3.3% as CFC-114a and ~3.3% as CFC-115, depending on the production route 22 . Estimates of global HFC-125 production increased from around 95 to 205 Gg yr −1 between 2010 and 2020 14 . These estimates of HFC production use the mean of the upper and lower estimates of a scenario of compliance with the Kigali Amendment 14 (Extended Data Fig. 5), which aims to phase out most HFC production, and is probably an underestimate of true HFC production for 2020. We use the estimated HFC-125 production and derived CFC emissions, minus their values in 2010, to approximate the ratio of CFC emissions to HFC production. Since 2010, CFC-115 emissions represent ~1.3%, CFC-113a emissions represent ~2.1% and CFC-114a emissions represent ~0.2% of HFC-125 production by mass. The timing and trajectory of the increase in CFC-113a and CFC-115 emissions follow those of HFC-125 production in countries under Article 5 of the Montreal Protocol (developing countries) since 2006 (Extended Data Fig. 5). Estimates of leakage rates for feedstock production vary substantially. A previous worst-case scenario for emissions due to by-product leakage during HFC production was 1.6% (ref. 23 ), yet typical leakage during fluorochemical production has also been estimated to be as high as 4% 24 (with a 95% uncertainty range of 0.1-20.0%). If undesirable by-products are formed during HFC-125 production, and not properly captured or destroyed, by-product leakage may explain at least some of the increasing emissions of CFC-113a, CFC-114a and CFC-115 over the past decade. The ratio of derived CFC emissions rates to HFC production supports the hypothesis of CFC leakage during HFC production.
In addition to being possible by-products in HFC-125 production, CFC-113a and CFC-114a are intermediates in a production pathway of HFC-134a. HFC-134a is the most abundant HFC in the atmosphere and is used primarily in mobile air conditioning, with estimated global consumption increasing from 237 to 321 Gg yr −1 between 2010 and 2020 14 (Extended Data Fig. 5). Emissions above 2010 levels for CFC-113a and CFC-114a are equivalent to around 3.5% and 0.3% of HFC-134a production, or 1.2% and 0.1% of combined HFC-125 and HFC-134a production. Leakage from feedstock usage of CFC-113a to produce trifluoroacetic acid, pesticides and the hydrofluoroolefin HFO-1336mzz may also contribute to the emissions 25 , but there are no published estimates of the produced quantities of these target substances. HFO-1336mzz will replace some HFCs with very high global warming potentials, and its production has the potential to continue long into the future. The hydrochlorofluorocarbon HCFC-133a is another undesirable by-product of HFC-125 production (up to 1.3% of perchloroethylene), and its emissions were previously attributed to HFC-125 and HFC-134a production 10,11,17,25 . Persistent emissions of HCFC-133a give supporting evidence for intermediate/by-product leakage during HFC production. However, emissions of HCFC-133a do not exhibit the same continuously increasing trend as the associated CFCs, which may suggest that not all sources of CFC and HCFC emissions are common during HFC production. In addition, emissions of HCFC-132b, also an intermediate/by-product of HFC-134a production, have increased from near zero in 2000 to around 1 Gg yr −1 in 2019 11 . Emissions of HCFC-31 have persisted, but have not increased, since 2010 11 and are associated with the production of HFC-32. HFC-32 is used with HFC-125 in the common refrigerator blend R-410A. Production of HFC-32 and HFC-125 for R-410A is positively correlated, although this correlation may not hold for production-related emissions of by-product gases.
If the Kigali Amendment to the Montreal Protocol had been universally ratified in 2016, production of HFC-134a was projected to be around 8% smaller in 2030 than in 2020 14 . To date, full ratification has not been achieved. However, even under this ideal scenario, HFC-125 production is still projected to increase by 18% between 2020 and 2030. If HFC-125 production alone is the source of CFC-113a, CFC-114a and CFC-115 emissions, and similar leakage rates continue, further increases in emissions of these CFCs are likely over the next decade even under universal ratification of the Kigali Amendment as HFC production continues to increase. Emissions of CFC-113a, CFC-114a and CFC-115 could reach around 4 ODP-Gg yr −1 in 2030, using average projected HFC-125 production assuming universal ratification of the Kigali Amendment 14 and our estimated leakage rates. If emissions of CFCs from feedstock use continue to rise, the ozone depletion associated with HFC and HFO production will continue to increase.

Uncertain drivers behind CFC-13 and CFC-112a
CFC-13 may be produced during the deliberate plasma arc destruction of CFC-12 refrigerants from old appliances 15 . Reported consumption of all CFCs to the United Nations Environment Programme has been negative since 2011 26 , meaning that CFCs are being destroyed globally. However, there is no information on the quantities of CFC-12 destroyed using plasma arc destruction versus other destruction methods, and thus the interconversion levels and eventual emissions of CFC-13 are unclear. Previous work has suggested that aluminium smelters are only a minor source of global emissions of CFC-13 9 . CFC-13 may be a by-product during the production of CFC-11 9 . Unreported production of CFC-11 occurred during the period 2013-2018 6 , and thus CFC-13 may have been inadvertently produced. However, unlike global emissions of CFC-11, which dropped rapidly after 2018 19 , emissions of CFC-13 continue through 2020, suggesting a different dominant source.
The source of the renewed emissions of CFC-112a is unknown and we can only speculate qualitatively on a possible resurgence of use as a solvent or a feedstock for fluorovinyl ether production 16 .

Importance for the Montreal Protocol
The trajectory of recently increasing emissions of CFC-13, CFC-112a, CFC-113a, CFC-114a and CFC-115 is counter to the goal of the production phase-out for dispersive uses, which was completed in 2010 ( Fig. 2 and Extended Data Fig. 4). The global annual rate of growth in the abundance of CFC-112a, CFC-114a and CFC-115 has accelerated rapidly since 2010. The rate of increase of the global abundance of CFC-113a has twice rapidly risen over the same period. Meanwhile, growth in the abundance of CFC-13 has continued to fluctuate since 2010 following an overall increasing trend. At present, the impact on stratospheric ozone is modest. Cumulative emissions of CFC-13, CFC-112a, CFC-113a, CFC-114a and CFC-115 between 2010 and 2020, relative to 2010 levels, could result in around a 0.002% loss in global stratospheric ozone, or 0.01% for the Antarctic Spring (following ref. 27 ). These emissions equate to an integrated ozone depletion of around 1.1 DU years 28 . However, if current trends continue or, as has already occurred twice for CFC-113a during 2010-2020, accelerate further, these influences would be larger and offset some of the progress made under the Montreal Protocol. Previous work has identified East Asia as a source of CFC-113a and CFC-115 emissions 8,9 . Updated regional emission estimates of CFC-13 and CFC-115 may better constrain the magnitudes of the increasing trends and their contributing emissions sources (CFC-13 and CFC-115 have wider measurement coverage than CFC-112a, CFC-113a and CFC-114a; Methods).
Ozone-depleting substances used as feedstocks and produced as by-products are not subject to the same controls on production as those for so-called dispersive use under the Montreal Protocol. As such, there is no current barrier to future use in the synthesis of chemicals. In the absence of further evidence, it is likely that the rapidly rising emissions of the long-lived ozone-depleting CFCs identified here are from processes not subject to current controls under the Montreal Protocol.

Online content
Any methods, additional references, Nature Portfolio reporting summaries, source data, extended data, supplementary information, acknowledgements, peer review information; details of author contributions and competing interests; and statements of data and code availability are available at https://doi.org/10.1038/s41561-023-01147-w.
Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law. On AGAGE and NOAA analytical systems, air sample measurements were bracketed by those of calibrated tanks of compressed whole air. Tanks from NOAA 30 and AGAGE 29 are calibrated against standards established through accepted gravimetrically derived methods. Cylinders from UEA/FZJ are calibrated against volumetric standards, where primary standards are based on a two-step static dilution in pure N 2 , which have been shown to compare well with gravimetrically derived calibration scales 10,17 . The calibration standards serve as reference links to the respective scales established at UEA (CFC-112, CFC-112a, CFC-113, CFC-113a, CFC-114 and CFC-114a) and NOAA (CFC-13 and CFC-115) and within AGAGE (CFC-115) and at the Swiss Federal Institute of Metrology (METAS; for CFC-13 AGAGE measurements).
At UEA/FZJ, air samples were analysed using 200-300 ml of air and systems based on cryogenic preconcentration at −78 °C followed by gas chromatograph coupled with mass spectrometer (GC-MS). UEA/FZJ secondary standards have been obtained from NOAA from air collected at Niwot Ridge, Colorado, USA (40.0542° N, 105.5889° W, 3,538 m above sea level). For more information on the measurement system, see refs. 8,18,31 . CFC-112, CFC-112a, CFC-113, CFC-113a, CFC-114 and CFC-114a measurements were performed on an Agilent 6890 GC coupled to a Waters AutoSpec triple-sector MS 8,17,18 . Two of the GC-MS systems were used for the analysis of the Cape Grim air archive 32 and flask samples in this work, one at UEA and one newly established at FZJ. The respective time series overlap for several years and show excellent agreement for all species, with similar uncertainties and no indications of any offset. Uncertainties are derived from the square root of the sum of the squared 1-sigma standard deviations of the working standard and the sample repeats, which were, on average, 4.1% (CFC-112a), 2.9% (CFC-113a) and 1.3% (CFC-114a) between 2000 and 2020. The different chromatographic set-ups of these analytical systems 17,18 enable the separation of CFC isomers, which is crucial for the derivation of isomer-independent trends and emissions.
AGAGE in situ measurements of CFC-13 and CFC-115 in ~2 l air samples were made using Medusa GC-MS instruments. These are cryogenic preconcentration systems coupled with a GC (Agilent) and a quadrupole mass selective detector (Agilent) 33,34 . The MSs were upgraded at several AGAGE sites in the 2011-2013 time frame, resulting in improved precision 9 for CFC-13 from ~1.5% (1 sigma) to ~1%. CFC-115 precision was ~0.8-0.4% (with only small improvements following the MS upgrades). Ambient air and reference gas measurements were alternated, resulting in ~12 fully calibrated samples per day 29 . The same kind of instruments were used to measure archived air samples (also typically ~2 l) collected as part of the Cape Grim Air Archive at Cape Grim 35 at the Commonwealth Scientific and Industrial Research Organisation and at the Scripps Institution of Oceanography (SIO) 36 . Measurements of CFC-13 are reported on the METAS-2017 primary calibration scale and measurements of CFC-115 on the SIO-05 scale as parts per trillion dry-air mole fractions 9 . Secondary and tertiary AGAGE calibration standards are compressed clean-air samples filled in the wintertime during stormy conditions at La Jolla (recent) or Trinidad Head (past), California, USA. Quaternary standards were filled either also at these sites or near individual measurement stations 29 . AGAGE measurements of archived air samples and in situ data are representative of both background and polluted air and therefore were filtered with statistical methods to remove outliers and pollution events to produce monthly means from the high-frequency measurements 37 .
The data provided from the NOAA network (Extended Data

Emission modelling
Emissions were estimated using measurements of dry-air mole fraction and a model of global atmospheric transport 40,41 . The estimation procedure follows a Bayesian approach modified from refs. 42,43 . The Bayesian approach involves inferring emissions from the measurements of dry-air mole fraction by means of an affine relationship between emissions and atmospheric mole fraction, considering the associated uncertainties, and a prior constraint on the emissions. The affine relationship is derived using the model of atmospheric transport.
Atmospheric transport is simulated using a 12-box model 44 , which divides the atmosphere into 4 semi-hemispheric zonal boxes (at 30° N and S) and vertically at 500 and 200 hPa (approximately representing the planetary boundary layer height and tropopause). Transport is seasonally averaged and repeats annually.
Measurements of dry-air mole fraction are averaged into monthly means, which are assumed to be representative of the semi-hemisphere surface box in which the measurement station is situated. Measurements from the AGAGE network were filtered to remove measurements Nature Geoscience Article https://doi.org/10.1038/s41561-023-01147-w not representative of background conditions using a statistical filter 37 before averaging.
A prior constraint is placed on the year-to-year growth in emissions, where the 1-sigma standard deviation in the growth is set equal to 20% of the mean a priori emissions over the inversion period (5% for CFC-114a). Due to the lack of information on emissions, a priori emissions of CFC-112a were somewhat arbitrarily assumed to be 0.05 Gg yr −1 each year. CFC-113a and CFC-114a were assumed to be 1 Gg yr −1 each year. Changing the a priori emissions by a factor of two had very little impact on the derived emissions (≲5% of the mean estimate) and does not change any conclusions drawn. CFC-115 a priori emissions were taken from ref. 9 (reaching a maximum of 12.7 Gg yr −1 in 1992), and CFC-13 a priori emissions were assumed to be one-seventh of CFC-115, following an earlier comparison of production estimates 9 .
Reported emissions are the maximum a posteriori estimates through minimization of a cost function 43 . In this work, the minimization used an iterative procedure (L-BFGS-B algorithm 45 ), which was constrained to ensure only positive emissions. Random uncertainties were calculated by assuming linearity around the maximum a posteriori estimate (Gaussian uncertainties), and additional systematic uncertainties were included due to errors in measurement calibration scales (3%), model transport error (1%) and atmospheric lifetime. Systematic uncertainties were found by random sampling the posterior covariance matrix, including the additional error terms. Lifetime errors are defined on the inverse lifetime and are assumed to be 20%, 55%, 80%, 20% and 20% for CFC-13, CFC-112a, CFC-113a, CFC-114a and CFC-115, respectively 17,46 .
Our work does not consider interannual changes to large-scale dynamics, which are known to impact emissions estimates of CFCs derived from atmospheric observations 19,47,48 . However, the sustained ten years of increasing emissions presented here far exceeds that of, for example, the mean period of the quasi-biennial oscillation of the direction of the zonal wind in the tropical lower stratosphere (~28 months 49 ). Errors in the year-to-year variability between emissions of CFCs may be introduced when changes to large-scale dynamical features of the atmosphere are neglected, but we are confident that the long-term trends are robust.

Derivation of global mean mole fractions and growth rates
Semi-hemispheric mole fractions were modelled using the derived emissions in each semi-hemisphere using the box-model approach described in Emission modelling and the affine relationship between emissions and mole fraction. The reported global mole fractions are the averages of the derived semi-hemispheric mole fractions for measurements from AGAGE and UEA/FZJ. Measurements of CFC-112, CFC-112a, CFC-113, CFC-113a, CFC-114 and CFC-114a were made only from air samples from Cape Grim in the southernmost semi-hemisphere. Therefore, derived global mole fractions in other semi-hemispheres are fully inferred from measurements in one semi-hemisphere. The additional uncertainty in the emission estimates arising from using measurements from only one semi-hemisphere is reflected by the uncertainty quantified using the Bayesian methodology. Such an approach has previously been used extensively to derive global emissions and mole fraction estimates of CFCs 8,17,18 . Mole fraction growth rates are calculated as an annual growth rate per month, which is smoothed over an approximately 18-month window using a Kolmogorov-Zurbenko filter 50 . The estimated global and hemispheric mole fractions from the NOAA record are weighted by the cosine of the latitude of the measurement site locations, from which the annual mean is taken.

Atmospheric lifetimes and metrics for ozone depletion and climate impact
The ODP is a metric to assess the ability of a gas to deplete ozone in the stratosphere and is calculated relative to that of CFC-11 (which has an ODP of 1). Both the ODP and atmospheric lifetimes of CFCs have been taken from ref. 12 as the most recent estimates. Global warming potentials over a 100 yr time horizon (GWP-100) have since been updated using newly available absorption spectra, revised atmospheric lifetimes 12 and methodology to calculate radiative efficiencies, and we refer to ref. 13 for these values. CFC-13 has an assumed ODP of 1.0 and is a potent greenhouse gas, with a GWP-100 17,200 times that of CO 2 . Its steady-state atmospheric lifetime is estimated at 640 years, with the dominant sink being reaction with O( 1 D). CFC-112a has an ODP of 0.86 and a GWP-100 of 3,740, with an atmospheric lifetime estimate of 52 years. The atmospheric sink of CFC-113a is due primarily to stratospheric photolysis, giving an atmospheric lifetime of 55 years, an ODP of 0.73 and a GWP-100 of 4,140. The primary O( 1 D) sink results in a CFC-114a atmospheric lifetime of 105 years, an ODP of 0.72 and a GWP-100 of 7,850. CFC-115 has an atmospheric lifetime of 540 years, an ODP of 0.26 and a GWP-100 of 10,200. The ODPs for CFC-112, CFC-113 and CFC-114 were estimated to be 0.98, 0.815 and 0.5, respectively, and GWPs were 4,880, 6,900 and 9,990, respectively. The integrated ozone depletion was calculated using an empirical formula 28 considering the total chlorine emissions and assuming that all loss occurs in the stratosphere.  52 , which also contains measurements analysed at UEA/FZJ.

Code Availability
The 12-box model and the inverse method used to quantify emissions are available via GitHub (https://github.com/mrghg/py12box (last accessed 16 November 2022) and https://github.com/mrghg/ py12box_invert (last accessed 16 November 2022)) and Zenodo (https:// doi.org/10.5281/zenodo.6857447 ref. 53   CFC-115 for the available direct atmospheric record (that is, excluding trends derived from firn air) from 1978 through 2020 (1999 through 2020 for CFC-112a). Circles show measured monthly mean mole fractions and lines are the model estimated mole fractions and growth rates, with the colours indicating the semi-hemispheric modelled estimate in which the measurement was made. Uncertainties (1 sd) are shown by the grey shading and are only shown for the global mean mole fractions and growth rates. Details on previously published records can be found in refs. 8,9,17,18 . CFC-13 abundances have increased continuously since 1978. The CFC-13 growth rate was rapid in the 1980s, slowed to 0.015 ppt yr −1 in 2000, and subsequently increased over the next two decades. From 2010 to 2020, CFC-13 increased by 9% from 3.04 ppt to 3.31 ppt (3.44 ppt in 2020 using the NOAA record, only available since 2014; see Extended Data Fig. 2). CFC-112a has the lowest abundance of the five CFCs. Its abundance was declining 1999-2006, but rapidly grew from 2010 to 2020 by 19% from 0.066 ppt to 0.078 ppt. The CFC-113a abundance grew since 1978 at a rate less than 0.02 ppt yr −1 until 2010, but from 2010 to 2020 it grew by 141% from 0.43 ppt to 1.02 ppt. CFC-114a growth rates were rapidly accelerating prior to 1990, then dropped to near zero between 1998-2003, after which they accelerated again. Between 2010 and 2020 CFC-114a increased by 9% from 1.03 ppt to 1.13 ppt. Similar to CFC-13 and CFC-114a, CFC-115 abundances also grew rapidly prior to the 1990s, but the timing is different. CFC-115 growth dropped to close to zero between 2005-2011, after which its growth rate again slowly increased. From 2010 to 2020, CFC-115 abundance grew by 4% from 8.38 ppt to 8.71 ppt (growing from 8.42 to 8.75 ppt from 2010 to 2020, or by 4%, using the NOAA record, Extended Data Fig. 2).

Extended Data Fig. 5 | Comparison of HFC production and CFC emissions
which may be associated with HFC production. The production of HFC-125 (a) and HFC-134a (b) is shown by the blue shading for Article 5 countries and orange shading for non-Article 5 countries. The composite of these two shading blocks shows total global production (without any metric of uncertainty). The estimates of HFC production assumes that all countries have ratified the Kigali Amendment, which to date would underestimate HFC production, and the values displayed are the mean of the lower and upper estimates from ref. 14