Unexpected nascent atmospheric emissions of three ozone-depleting hydrochlorofluorocarbons

Significance

We demonstrate the need to detect and track unexpected substances in the atmosphere and to locate their sources. Here, we report on three hydrochlorofluorocarbons (HCFCs) that have no known end-uses. HCFC-132b (${\mathrm{C}\mathrm{H}}_2{\mathrm{C}\mathrm{l}\mathrm{C}\mathrm{C}\mathrm{l}\mathrm{F}}_2$) is newly discovered in the global atmosphere. We identify East Asia as the dominant source region for global emissions of this compound and of HCFC-133a (${\mathrm{C}\mathrm{H}}_2{\mathrm{C}\mathrm{l}\mathrm{C}\mathrm{F}}_3$). We also quantify global emissions of HCFC-31 (${\mathrm{C}\mathrm{H}}_2$ClF). These compounds are most likely emitted as intermediate by-products of chemical production processes. The early discovery and identification of such unexpected emissions can identify the related industrial practices and help to develop and manage environmental policies to reduce unwanted and potentially harmful emissions before the scale of the problem becomes more costly to mitigate.

Abstract

Global and regional atmospheric measurements and modeling can play key roles in discovering and quantifying unexpected nascent emissions of environmentally important substances. We focus here on three hydrochlorofluorocarbons (HCFCs) that are restricted by the Montreal Protocol because of their roles in stratospheric ozone depletion. Based on measurements of archived air samples and on in situ measurements at stations of the Advanced Global Atmospheric Gases Experiment (AGAGE) network, we report global abundances, trends, and regional enhancements for HCFC-132b (${\mathrm{C}\mathrm{H}}_2{\mathrm{C}\mathrm{l}\mathrm{C}\mathrm{C}\mathrm{l}\mathrm{F}}_2$), which is newly discovered in the atmosphere, and updated results for HCFC-133a (${\mathrm{C}\mathrm{H}}_2{\mathrm{C}\mathrm{l}\mathrm{C}\mathrm{F}}_3$) and HCFC-31 (${\mathrm{C}\mathrm{H}}_2$ClF). No purposeful end-use is known for any of these compounds. We find that HCFC-132b appeared in the atmosphere 20 y ago and that its global emissions increased to 1.1 Gg$\cdot$y⁻¹ by 2019. Regional top-down emission estimates for East Asia, based on high-frequency measurements for 2016–2019, account for $\sim$95% of the global HCFC-132b emissions and for $\sim$80% of the global HCFC-133a emissions of 2.3 Gg$\cdot$y⁻¹ during this period. Global emissions of HCFC-31 for the same period are 0.71 Gg$\cdot$y⁻¹. Small European emissions of HCFC-132b and HCFC-133a, found in southeastern France, ceased in early 2017 when a fluorocarbon production facility in that area closed. Although unreported emissive end-uses cannot be ruled out, all three compounds are most likely emitted as intermediate by-products in chemical production pathways. Identification of harmful emissions to the atmosphere at an early stage can guide the effective development of global and regional environmental policy.

Localizing and quantifying halocarbon emissions from atmospheric observations and transport modeling has become an important tool to validate emissions derived from activity data and emission factors (1–7). This can also be used to detect new substances and derive their nascent trends and emissions, thereby playing an important role as an early warning system leading to improved environmental emissions policies.

Here, we present long-term emissions of three ozone-depleting substances (ODSs) which have no reported end-use. The emissive use of these substances is regulated by the Montreal Protocol on Substances that Deplete the Ozone Layer and its Amendments (hereafter referred to as the Montreal Protocol). The Montreal Protocol is an international agreement that regulates the phase-out of production and consumption of ODSs. The environmental target of these regulations is to lower ODS abundances in the atmosphere to safeguard the stratospheric ozone layer. The full ban on production and consumption for emissive end-use of the primary ODSs, the chlorofluorocarbons (CFCs), was set to the mid-1990s for developed (non-Article 5) countries and to 2010 globally. As a consequence, emissions have been declining when calculated based on production and consumption (bottom-up). Unexpectedly, though, emissions inferred from atmospheric observations (top-down method) of several ODSs were recently found to be declining more slowly than expected, or even increasing (4, 5, 8–10). This raised concerns about potential violations of the Montreal Protocol (4, 5, 11). However, it is difficult to prove a violation, because emissions are aggregated when using the top-down method, and additionally include those from banked ODSs in end-user products (e.g., refrigerators or foams), which are not controlled by the Montreal Protocol (12). Further, emissions from feedstock and process agents, and from inadvertent or coincidental production during a manufacturing process, are also included in these aggregated top-down emission estimates. Although the Montreal Protocol addresses this group of emissions, no stringent control is enforced to date, with the parties primarily being urged to take steps to minimize such emissions (11, 13, 14). Attention has so far been limited to only a small number of compounds, notably carbon tetrachloride (${\mathrm{C}\mathrm{C}\mathrm{l}}_4$), whose emissions are not declining as expected, partially due to unreported nonfeedstock emissions (3, 15).

The three ODSs identified in this study are all hydrochlorofluorocarbons (HCFCs), which have lower potentials than CFCs to harm the ozone layer, and which have been used in the past as interim replacements for CFCs. Their phase-out by the Montreal Protocol was significantly tightened in 2007, with a complete ban in 2020 for developed (non-Article 5) countries and 2030 globally. HCFCs have also been included in the baseline calculations under the Kigali Amendment in 2016 to facilitate “leap-frogging” high global warming potential (GWP) hydrofluorocarbons (HFCs) by directly replacing HCFCs with low-GWP substances (16, 17). Similar to the CFCs, the major HCFCs, HCFC-22 (${\mathrm{C}\mathrm{H}\mathrm{C}\mathrm{l}\mathrm{F}}_2$), HCFC-142b (${\mathrm{C}\mathrm{H}}_3{\mathrm{C}\mathrm{C}\mathrm{l}\mathrm{F}}_2$), and HCFC-141b (${\mathrm{C}\mathrm{H}}_3{\mathrm{C}\mathrm{C}\mathrm{l}}_2$F), are used in stationary refrigeration and structural foam blowing. Their global emissions have leveled or started to decline over the past years as a consequence of their production phase-down (18).

We report on the newly detected HCFC-132b (1,2-dichloro-1,1-difluoroethane; ${\mathrm{C}\mathrm{H}}_2{\mathrm{C}\mathrm{l}\mathrm{C}\mathrm{C}\mathrm{l}\mathrm{F}}_2$) in the atmosphere and present substantial updates on abundance and emissions for the previously found HCFC-133a and HCFC-31 (8, 19, 20). Their lack of known end-uses gives rise to speculation about their sources and their roles within the framework of the Montreal Protocol. There are no public inventories or bottom-up emission reports available for these compounds. Although their physical and chemical properties are suitable for applications in refrigeration and other industrial applications, their toxicities and carcinogenicities have prevented consumer end-use applications in the past (21–25). Their removal from the atmosphere is mainly driven by reaction with the hydroxyl radical (OH), leading to global atmospheric lifetimes of 3.5 y for HCFC-132b, 4.6 y for HCFC-133a, and 1.2 y for HCFC-31 (26).

The measurements presented here have wide geographical and temporal coverage, based on ongoing in situ ground-based measurements at the stations of the Advanced Global Atmospheric Gases Experiment (AGAGE) network (27). Our records also include measurements of archived air samples from the Southern Hemisphere (SH) Cape Grim Air Archive (CGAA) starting in 1978 and from the Northern Hemisphere (NH), as well as multiyear weekly collected air samples from Antarctica. We also extend the previous record of HCFC-133a with in situ AGAGE and new CGAA measurements (8, 19). Furthermore, we present a longer record for HCFC-31 than was previously available (20), through measurements of the CGAA, and updated contemporary observations from Antarctic samples (2015–2019) and NH measurements from Dübendorf (Switzerland).

Using these measurements, an inverse method, and the AGAGE 12-box atmospheric transport model, we estimate hemispheric emissions and reconstruct abundances of the three HCFCs from 1978 to the present (28, 29). Based on the large pollution events recorded at the AGAGE station, Gosan (Jeju Island, South Korea), we also estimate regional East Asian emissions of HCFC-132b and HCFC-133a using a regional inverse-modeling system (30).

Results

Global Atmospheric Distributions of the HCFCs.

We calculated hemispheric long-term trends by combining flask measurements with in situ high-resolution observations from AGAGE stations (Fig. 1). HCFC-132b first appeared in the NH atmosphere in the late 1990s, followed by a sustained and rapid growth to a dry air mole fraction of 0.15 parts per trillion (ppt) by 2013 (Fig. 1A). After a short decline until 2016, the compound increased again to a maximum of 0.17 ppt by the end of 2019. The SH abundances lagged the NH abundances and remained lower throughout the entire record, indicating that emissions of this compound predominantly occurred in the NH. The absence of HCFC-132b from the atmosphere before 1995 (SI Appendix) suggests an entirely anthropogenic origin.

Fig. 1.

Atmospheric observations and model reconstructions of the HCFCs, HCFC-132b (A), HCFC-133a (B), and HCFC-31 (C). Units are dry air mole fractions in ppt (pmol$\cdot {\mathrm{m}\mathrm{o}\mathrm{l}}^{-1}$). The $\sim$40-y-long records are limited here to 1999–2019 for better temporal resolution. Flask-sample results from various NH sites are aggregated into a single dataset for clarity. Flask samples for the SH are shown for Cape Grim (CGAA, Tasmania) and the Antarctic station King Sejong. In situ records for HCFC-132b and HCFC-133a from stations of the AGAGE are shown here as background-filtered monthly means for a few illustrative stations only. Modeled records, derived from the observations and an inversion system using a 12-box chemical transport model, are shown for the four ground-level model boxes.

HCFC-133a exhibits a general increase in both hemispheres. Measurements of archived air from the US West Coast detail a pronounced reversal in the NH abundance in 2007/2008, in agreement with a similar feature found for the SH (8). Also, independent measurements of the CGAA confirm earlier findings, in particular, the presence of this compound in the SH atmosphere before 1978 (8). New flask and in situ measurements for 2015–2019 reveal that the downward trend of HCFC-133a in the NH (2012–2015; ref. 19) has reversed, and the compound has increased to >0.5 ppt again.

For HCFC-31, first detectable mole fractions appeared in samples from the late 1990s. Following more than a decade-long growth, we find, similar to HCFC-133a, a decline of HCFC-31 in the atmosphere for 2012–2015, which was followed by another strong increase and a stabilization over the past 3 y.

There are surprising similarities in the records of the three compounds (Fig. 1). Compared to known records for many other halocarbons, we find large multiyear variability in these records, pointing to rapidly changing emissions. The most pronounced feature is a temporal maximum in abundances for all three compounds within the period 2012–2014, although not exactly synchronous.

Global Emissions.

Global emissions for all three compounds show a generally increasing trend over the last two decades, with mean values for 2016–2019 of 0.97 Gg$\cdot$y⁻¹ for HCFC-132b, 2.3 Gg$\cdot$y⁻¹ for HCFC-133a, and 0.71 Gg$\cdot$y⁻¹ for HCFC-31 (Fig. 2). However, we calculated a large relative variability in these emissions, particularly for HCFC-133a. This variability is unusual compared to other widely used synthetic halocarbons (18) and indicates that a major fraction of these emissions does not originate from banks (compounds stored in equipment, which are usually emitted slowly over time and, hence, only exhibit small variability in their global emissions). It further suggests that the emissions are not deriving from impurities in commercially used halocarbons, which generally show temporally much smoother emission trends. Note that the emissions uncertainties in Fig. 2 are dominated by uncertainties in the lifetime, which act like potential biases across all years. Therefore, in our global inversions, the year-to-year variability is better constrained than the absolute emissions magnitude, particularly for HCFC-132b, which has a relatively large lifetime uncertainty.

Fig. 2.

Global and East Asia regional emissions of the HCFCs HCFC-132b (A), HCFC-133a (B), and HCFC-31 (C). Shaded gray bands denote the 16/84-percentile uncertainty range for the global emissions. Total emissions from East Asia (orange) and Eastern China only (maroon) are plotted for HCFC-132b and HCFC-133a with 95% confidence intervals (gray shading).

Twenty-year cumulative emissions (1978–2019) for HCFC-132b, HCFC-133a, and HCFC-31 amount to 13 Gg, 44 Gg, and 10.6 Gg, respectively. Given their relatively small ozone-depletion potentials (ODPs), compared to the primary ODSs, of 0.038, 0.019, and 0.019, respectively, (26), we calculated a combined cumulative emission of 1.5 Gg-ODP-weighted. Although their impacts on stratospheric ozone degradation are small and roughly one order of magnitude smaller compared to the recently found yearly unexpected emissions of CFC-11 (${\mathrm{C}\mathrm{C}\mathrm{l}}_3$F) (4, 5), their absolute emissions are significant, particularly for compounds with no purposeful end-use.

Emissions from East Asia.

Within the AGAGE network, frequent and large (up to 4 ppt) pollution events for HCFC-132b and HCFC-133a were recorded at the South Korean station, Gosan, indicating substantial regional emissions (Fig. 3; HCFC-31 is not measured at the AGAGE sites; Measurement Methods). By combining these records with an inverse-modeling method (30), we find that the most concentrated emissions in East Asia (defined as China, Taiwan, North and South Korea, and Japan) occur in Eastern China (Figs. 2 and 4). For HCFC-132b, Eastern China emissions are 0.43 to 0.53 Gg$\cdot$y⁻¹ for 2016–2019 and account, on average, for 50% of global emissions (Fig. 2A). The East Asia total emissions account for $\sim$95% of the global emissions, within the uncertainties of the methods. The inversion also attributes a large fraction of East Asian emissions to West China. However, due to the reduced sensitivity of the observational site to West China, these estimates are connected with a much larger uncertainty than for East China. For HCFC-133a, Eastern China emissions account, on average, for 43% of the global emissions, and East Asia emissions explain $\sim$80% of the global emissions (Fig. 2B).

Fig. 3.

High-resolution measurement records of HCFC-132b (A) and HCFC-133a (B) from selected stations of the AGAGE. Pollution events recorded at the Gosan station (South Korea), where measurements started in 2015/2016, strongly exceed those at all other stations in frequency and magnitude.

Fig. 4.

Posterior HCFC-132b (A) and HCFC-133a (B) emission distributions ($\mathrm{\mu }$g$\cdot {\mathrm{k}\mathrm{m}}^{-2}\cdot {\mathrm{s}}^{-1}$) for eastern China based on observations at Gosan (Jeju Island, South Korea) for 2016–2019. The red line encloses the area defined as East China.

There is a distinct difference in the geographical distribution of the emissions from Eastern China (Fig. 4). For HCFC-132b, the strongest source region is found in northeastern China (Shangdong and Southern Hebei). In contrast, for HCFC-133a, the highest emissions are found in the Shanghai region. Both regions were recently identified as strong emitters of other halocarbons, but the predominance of HCFC-132b and HCFC-133a emissions to only one of these two regions is unusual (10, 31). Both areas host intense fluorocarbon industry, which could support speculations on feedstock/by-product emissions. For ${\mathrm{C}\mathrm{C}\mathrm{l}}_4$, another ODS with little-known allowed emissive end-uses, a study covering 2009–2016 found that emissions first originated in the Shanghai region, but then spread out to include northern Chinese provinces (32). Recent similar high-emission regions were also found for CFC-11 with suggested ultimate emissive end-use (5).

Sources in Western Europe.

We find much smaller and highly sporadic (two to three times per year) pollution events for HCFC-132b (up to 0.5 ppt) and HCFC-133a (up to 3.5 ppt) at some of the European stations (mainly Jungfraujoch and Monte Cimone). Surprisingly, European pollution events for HCFC-132b ceased by early 2017, and those for HCFC-133a became even less frequent (SI Appendix), indicating that regional emissions were greatly reduced. For the period before April 2017, our analysis reveals strong emissions of HCFC-133a near Lyon in southeastern France, similar to those found earlier using a reduced dataset (19), and weaker HCFC-132b emissions close by. The emissions in this region of intense fluorochemical production ceased after April 2017, and only a secondary potential source of HCFC-133a remains in western Germany. The cessation of measured HCFC-132b and HCFC-133a pollution events in early 2017 could indicate a change in the manufacturing processes in relevant production plants in the area. A possible explanation is the cessation of HFC-134a (${\mathrm{C}\mathrm{H}}_2{\mathrm{F}\mathrm{C}\mathrm{F}}_3$) production at Pierre-Bénite (Lyon) in the first quarter of 2017 (33).

Discussion

HCFC-133a and HCFC-31 in the global atmosphere were previously assumed to originate from factory-level emissions during the production of mainly HFC-134a and HFC-143a (${\text{CH}}_3$${\text{CF}}_3$) and HFC-32 (CHCl₂F), respectively (8, 14, 20, 34, 35). HCFC-132b is also likely an intermediate/by-product involved in reactions to produce HFC-134a and perhaps other HFCs, though we cannot exclude end-use applications. For example, in the widely used reaction of trichloroethylene (CHCl=${\mathrm{C}\mathrm{C}\mathrm{l}}_2$) with hydrogen fluoride (HF) to produce HFCs (foremost HFC-134a), the intermediates ${\mathrm{C}\mathrm{H}}_2$Cl-${\mathrm{C}\mathrm{F}\mathrm{C}\mathrm{l}}_2$ (HCFC-131a), ${\mathrm{C}\mathrm{H}}_2$Cl-${\mathrm{C}\mathrm{C}\mathrm{l}\mathrm{F}}_2$ (HCFC-132b), and ${\mathrm{C}\mathrm{H}}_2$Cl-${\mathrm{C}\mathrm{F}}_3$ (HCFC-133a) could potentially be produced and leak to the atmosphere (36). The other isomers of dichlorodifluoroethane are not likely produced from hydrofluorination of HCFC-131a, as this would require hydrogen (H) rearrangement, in the case of CHClF-CHClF (HCFC-132) and ${\mathrm{C}\mathrm{H}\mathrm{C}\mathrm{l}}_2$-${\mathrm{C}\mathrm{H}\mathrm{F}}_2$ (HCFC-132a), or nucleophilic substitution of chlorine (Cl) by fluorine (F) on the less favorable carbon atom in the case of ${\mathrm{C}\mathrm{H}}_2$F-${\mathrm{C}\mathrm{C}\mathrm{l}}_2$F (HCFC-132c). This may explain why we did not find these three compounds in the global atmosphere.

Our finding of a strong source of HCFC-133a (and, to a lesser extent, also of HCFC-132b) in France potentially supports this hypothesis, in particular, given the lack of recorded pollution events for both compounds, starting with the cessation of the HFC-134a production in that region. On the other hand, for the other (and only remaining) European HFC-134a production site near Frankfurt, HCFC-133a emissions are detected with our method, while those for HCFC-132b are arguably undetectable (Fig. 5 and SI Appendix).

Fig. 5.

Potential source areas for European emissions of HCFC-132b and HCFC-133a, as estimated by footprint statistics (SI Appendix). Units refer to average mole fraction enhancements at the receptor sites when influenced by a given grid cell. A and C are for 2014–March 2017, and B and D are for April 2017–2019. This temporal distinction is made based on the lack of significant pollution events after March 2017. Stations are Mace Head (MHD; Ireland), Jungfraujoch (JFJ; Switzerland), and Monte Cimone (CMN ; Italy). Triangles denote HFC-134a factories in France (upward triangle) and Germany (downward triangle). Dotted areas have low source sensitivities and should not be interpreted.

For the globally dominant emissions, which we locate in East Asia, the predominant source regions for HCFC-132b and HCFC-133a are located in different places. Assuming that both substances are emitted mainly during HFC-134a production, this separation and the large temporal global emission variability, despite monotonically increasing global HFC-134a production, suggest that the generation and leaks of HCFC-132b and HCFC-133a (and HCFC-31 in the case of HFC-32 production) are highly sensitive to industrial practices at the individual HFC production facilities.

Conclusions and Significance

In addition to targeting ODS end-use applications, the Montreal Protocol also addresses feedstock and process emissions, however, with currently no stringent control. The need to place adequate emphasis on these emissions is demonstrated by the example of ${\mathrm{C}\mathrm{C}\mathrm{l}}_4$, a compound for which large global unaccounted emissions are found, a large fraction of which are unreported and believed to derive from current industrial production processes (3, 15). We report on emissions of three other ODSs which most likely fall into this category. Our findings of geographical source separations and large temporal variabilities in the global emissions suggest that some factories temporarily emit much more process-intermediate HCFCs than the low percentages of the end-use compound (HFC) that are commonly assumed (14). These should be identified, and measures applied for emission reductions, according to the recommendations of the Montreal Protocol.

While the emissions of these three HCFCs are quantifiable and have increased over the last decades, their ODP-weighted impacts are small compared to those of the major ODSs. Nevertheless, our findings demonstrate a method for early warning detection and quantification of nascent emissions of synthetic trace gases based on atmospheric observations. This can, in turn, help to validate inventory-based emissions estimates, uncover potential unreported sources, and enable the assessment of the effectiveness of subsequent mitigation efforts. From an economic perspective, early discovery also enables processes to be changed to reduce unwanted emissions before the scale of the problem becomes more costly to mitigate, as has unfortunately been the case for unaccounted emissions of other compounds (1–7).

More broadly, this study demonstrates the analytical power of modern atmospheric monitoring instrumentation to detect atmospheric abundances at sub-ppt level arising from nascent industrial emissions. It also demonstrates the importance of large-scale global atmospheric network observations and modeling for identifying and quantifying regional emissions. As industrialization continues to expand and move into new regions, the absence of broader regional coverage for such atmospheric observations inhibits a full reconciliation between emissions measured in the global atmosphere and the sum of those determined regionally. The expansion of observational networks will be needed to close such gaps in support of the protection of the stratospheric ozone layer and the climate.

Materials and Methods

Site Description.

The AGAGE global network consists of nine fully intercalibrated field stations with long-term measurement records of a suite of ODSs (27). The network is complemented by affiliated stations and by laboratory instruments, which also serve as urban stations. European field stations are Zeppelin (Spitsbergen), Mace Head (Ireland), Jungfraujoch (Swiss Alps), and the affiliated station Monte Cimone (Italy). Asian stations are located at Gosan (Jeju Island, South Korea) and Shangdianzi (China); however, data from the latter could not be used for this study. Other stations are Trinidad Head (California), Ragged Point (Barbados), Cape Matatula (American Samoa), and Cape Grim (Tasmania, Australia). Measurement records for HCFC-132b and HCFC-133a are of various lengths, with the longest at Jungfraujoch, starting in 2013. In addition to the in situ measurements, flask samples contribute to the analysis presented here. Canister samples are collected weekly at the South Korean Antarctic Station King Sejong (King George Island, South Shetland Islands) (37). Archived air samples used in our analysis derive from the CGAA, which is a set of >120 samples collected since 1978 at the Cape Grim Baseline Air Pollution Station (38–41). Additional archived air samples were collected under clean-air conditions in the NH, mostly at Trinidad Head, La Jolla (California), Boulder (Colorado), and in the Swiss Alps (42).

Measurement Methods.

All measurements presented here were conducted using “Medusa” preconcentration gas chromatography mass spectrometry (GC-MS) instrumentation used in AGAGE (27, 43, 44). HCFC-132b was identified along with the other three isomers of dichlorodifluoroethane to conclusively demonstrate the absence of interferences. Also, with first weekly atmospheric measurements of these three additional isomers, starting in mid-2019, and detection levels of $\sim$0.005 ppt, we find these compounds undetectable within the airmass footprint of the urban station Dübendorf (Switzerland). Due to their atmospheric lifetimes $>$1 y (17), we conclude that these three isomers are currently also undetectable in the global atmosphere.

Analytical details for HCFC-133a and HCFC-31 are given in earlier studies (19, 20). Since then, HCFC-133a measurements have been fully integrated into the AGAGE network, from which global high-resolution data are now available. Analysis of HCFC-31 is hampered by a coelution with CFC-12 on aged Porabond Q columns (used in AGAGE) and has therefore not been integrated into the network. The HCFC-31 flask-sample measurements used in this study were made on the Empa laboratory Medusa-GC-MS in a batch mode using a column on which the two compounds could be fully separated and on a GasPro column on the Commonwealth Scientific and Industrial Research Organization (CSIRO) laboratory Medusa (Medusa-9) for the CGAA samples.

Calibration.

The present work prompted the development of the Swiss Federal Office of Metrology (METAS) METAS-2017 primary calibration scale for HCFC-132b (45). It is based on 11 dynamically gravimetrically prepared, Système International-traceable primary reference standards ranging from 0.9 to 1.5 ppt. The calibration scale was adopted into the AGAGE-based Scripps Institution of Oceanography (SIO) R1 calibration-measurement database and allowed for a reporting of fully intercalibrated measurements used here. The estimated accuracy of this calibration scale is 1.6% (2$\sigma$). Measurements of HCFC-133a are also fully intercalibrated and are based on the Empa-2013 primary calibration scale with an estimated accuracy of 10% (2$\sigma$) (19). For HCFC-132b and HCFC-133a, the calibration scales were propagated from the SIO pool of secondary standards through tertiary traveling standards to on-site quaternary (working) standard. For HCFC-31, the Empa and CSIRO laboratory measurements were intercalibrated, and results are based on the Empa-2013 primary calibration scale (20).

Global Emissions Derived from Global Chemical Transport Model and Inverse Methods.

To derive global emissions that are based on baseline atmospheric observations (“top-down”), we employed the AGAGE 12-box model (46–50). The model divides the atmosphere into four zonal bands, separated at the Equator and at the $\mathrm{30}°$ latitudes, thereby creating boxes of similar air masses. The vertical box separations are at 500 and 200 hPa. We included temperature-dependent hydroxyl (OH) radical reactions, which are the main removal mechanism of all three compounds from the atmosphere (51). We assumed stratospheric lifetimes of 45 y for HCFC-132b, 103 y for HCFC-133a, and 35 y for HCFC-31 (26), leading to overall atmospheric lifetimes of 3.4 y for HCFC-132b, 4.5 y for HCFC-133a, and 1.4 y for HCFC-31 in the box model.

Emissions were inferred by comparing model simulations to baseline observations using a Bayesian inverse method in which the emissions growth rate was weakly constrained a priori (29). High-frequency observations were filtered to remove “pollution events” using a statistical filter (28) and, combined with archived air samples, averaged into semihemispheres. The uncertainty on the observations combined two terms, one related to the measurement repeatability and one related to the ability of the model to represent the data. The latter was assumed to be equal to the baseline variability for high-frequency samples (or, for flask samples, the average high-frequency variability in the same hemisphere, scaled by mole fraction difference between the flask and high-frequency observations). Some seasonality in emissions was imposed by fitting a sine curve to the emissions, which minimized the model–data mismatch (19). Given the lack of available prior emissions estimates, the annual emissions growth rate was assumed to be zero a priori, with an uncertainty assumed to be 0.1, 1, and 0.1 Gg$\cdot$y⁻² for HCFC-132b, HCFC-133a, and HCFC-31, respectively. The derived emissions were found to not strongly depend on these values. The uncertainties in the a posteriori emissions estimates combined uncertainties related to the measurements, model representation error, prior constraint, atmospheric lifetime, and calibration-scale uncertainty (29). The uncertainty in the lifetime was assumed to be 20% for HCFC-31 (20). For HCFC-133a and HCFC-132b, we assumed lifetime uncertainties equal to the uncertainty in the rate constant with respect to reaction with the hydroxyl radical [10% for HCFC-133a and 50% for HCFC-132b (51)], which we assumed to be the largest term in the lifetime uncertainty budget for these substances.

Regional Emission Estimates.

The inversion system used to estimate East Asian regional emissions has been documented (5, 30). It is here applied to HCFC-132b and HCFC-133a for 2016–2019 and using observations from Gosan alone, for which pollution events were identified and quantified by subtracting a smooth statistical baseline fit from the observations (52). Source sensitivities were computed with the Lagrangian particle-dispersion model FLEXPART (version 9.2-Empa) driven by operational analysis and forecasts from the European Center for Medium-Range Weather Forecasts Integrated Forecasting System employing a horizontal resolution of 0.$\mathrm{2}°\times 0.\mathrm{2}°$ in the area of interest. The Bayesian inversion for East Asia was carried out by using a flat a priori emission distribution over all land areas, reflecting our expectation that the emissions are not end-user-related and, hence, don’t follow population densities. A priori and data-mismatch covariance matrices were constructed by using a maximum-likelihood approach (30). For European HCFC emissions, the inversion system showed very little skill in reproducing the observed peak concentrations and was consistent with emissions originating from intermittent sources. Therefore, we applied a more qualitative method to identify potential source areas, which can be seen as a spatially distributed, weighted averaging of the observed concentration increments and builds on a method termed “trajectory statistics” (53) (SI Appendix).

Data Availability.

In situ measurements of HCFC-132b and HCFC-133a from the AGAGE stations are available through the AGAGE website (https://agage.mit.edu/). Measurements of the samples collected in flasks, such as for archived air, Antarctica and Dübendorf (HCFC-31), and model results, are accessible through https://zenodo.org (DOI: 10.5281/zenodo.4266485).