Continuing industrial emissions are delaying the recovery of the stratospheric ozone layer

Abstract

The Montreal Protocol on Substances that Deplete the Ozone Layer has greatly restricted the global production and consumption of long-lived ozone-depleting substances (ODS). However, ODS used or consumed as feedstocks in the manufacture of other chemicals are excluded from restrictions. This exclusion was based on the assumption that emission rates of feedstocks were only 0.5% of the amount produced and that feedstock production would decline in the future, with remaining emissions too small to significantly affect the stratospheric ozone or its recovery. In sharp contrast, feedstock emissions are now assessed as being substantially higher (typically 3.6% of production), and feedstock production and use has been rising rather than falling. Here, scenarios in which feedstock-related ODS emissions continue at this current fraction of production, rather than the 0.5% reference case, are projected until 2100. Without additional measures, these elevated emissions could delay the recovery of the mid-latitude stratospheric ozone layer by 7 (6 – 11) years. Furthermore, limiting ODS feedstock emissions would also reduce their effect on direct radiative forcing and on climate change.

Introduction

Under the 1987 Montreal Protocol and its amendments, the production and consumption of ODS with a high ozone-depletion potential (ODP) have been globally banned for emissive uses since 2010. This group of ODS includes chlorofluorocarbons (CFCs) and other long-lived chlorinated and brominated compounds. Hydrochlorofluorocarbons (HCFCs) were introduced as interim substitutes for ODS with a high ODP. They are also ODS but due to their smaller effect on ozone destruction, production for emissive uses is still permitted in very limited quantities until 2040 in developing countries defined under the Montreal Protocol.

In contrast, the production and consumption of ODS for feedstock applications (i.e. as chemical building blocks in the production of other chemical end products) has no restrictions under the Protocol. However, under Article 7 of the Protocol, countries are obliged to report their annual feedstock production and consumption to the United Nations Environment Programme (UNEP).

In the 1990 London Amendment, Parties were additionally urged to take steps to minimize emissions related to feedstocks. Consequently, the Technology and Economic Assessment Panel of the Montreal Protocol (TEAP) was requested to provide an estimate of feedstock-related emissions and reported that actual emissions from existing well-designed and managed facilities are estimated to be about 0.5% of the production¹. Furthermore, it was projected that total emissions from feedstock production would decrease between the mid-1990s and 2000². These evaluations were carried out just before the phase-out of CFC-11 and CFC-12, for which carbon tetrachloride (CCl₄) was a major feedstock. At that time, other ODS were not expected to have a large importance as feedstocks in later years. This contributed to the perception that the impact of emissions from feedstocks, as intermediates, and as by-products, would be negligible and would remain a minor concern in the future. If the assumptions that justified the exclusion of feedstock ODS under the Montreal Protocol had remained accurate, emissions of ODS after 2010 would have stemmed almost entirely from the slowly declining release of ODS from banks (e.g. refrigerants, foam-blowing agents, and other contained applications) and remaining uses of HCFCs.

To analyse the feedstock chemicals and their main applications, a detailed map of the large diversity of currently known, major industrial processes that involve ODS as feedstocks, intermediates and by-products is shown in Fig. 1. Nowadays, ODS feedstocks are used in large quantities to produce non-ozone-depleting hydrofluorocarbons (HFCs), which are currently being phased-down following the 2016 Kigali Amendment to the Montreal Protocol. They are also used to produce short-lived hydrofluoroolefins (HFOs) and hydrochlorofluoroolefins (HCFOs), which are unsaturated fluorochemicals not controlled under the Montreal Protocol, used as substitutes for other controlled gases. Furthermore, substantial quantities of ODS feedstocks are used to produce halogenated polymers and additional halogenated chemicals (e.g. polytetrafluoroethylene (PTFE), trifluoroacetic acid, and fipronil—an insecticide). Many of these end products were produced in much smaller quantities, or not at all, during the early phases of the Montreal Protocol, resulting in no restrictions being imposed on emissions of feedstock-related processes until today.

Fig. 1: Use of controlled ODS as feedstocks.

Feedstock use of controlled ODS (ozone-depleting substances) under the Montreal Protocol (red) and unrestricted chemicals (blue) to produce controlled hydrofluorocarbons (HFCs, green), uncontrolled hydro(chloro)fluoroolefins (H(C)FOs, orange), other chemicals (grey), and polymers (purple). Paths leading to intended end products are shown as solid lines, while those leading to by-products are shown as dashed lines. HFC-23, which is mostly a by-product, and for which destruction is prescribed to the extent practicable under the Kigali Amendment, is separated from the other HFCs. The production of HFCs, addressed within the phase down under the Kigali Amendment, is assessed to decline accordingly²⁶. For all other products, their expected annual increase in production is shown, as discussed in Table 1 and in the Supplementary Information.

In addition, various ODS are also formed as unavoidable by-products in the manufacture of specific chemicals. Examples include CCl₄, which is generated during the production of polyvinyl chloride (PVC), as well as CFC-115 and CFC-114/a (CFC-114 and CFC-114a collectively), which are by-products of HFC-125 production. These by-products are not further utilised in any current production chain and could potentially be separated and destroyed. Figure 1 also shows intermediates (e.g. HCFC-132b) that should be completely transformed into other substances at a single location, and which should not be released outside the production chain. These substances are therefore defined as intermediates rather than as feedstocks. However, all feedstocks, by-products, and intermediates in production processes are potential emission sources because of their unavoidable leakage to the atmosphere, and they are hereafter collectively referred to as feedstock emissions, unless otherwise indicated. In addition to ODS, certain HFCs are also used as feedstocks (i.e. HFC-23, HFC-152a, HFC-245fa), which, for completeness, are included in Fig. 1.

Results and discussion

Higher than expected emission factors and increasing use of ODS feedstocks

Despite earlier projections, the reported global use of ODS as feedstock chemicals did not decline after 2000³ but has increased by 163% between 2000 and 2024 (Fig. S1). The highest growth between 2014 and 2024 was for HCFC-22, CCl_4, CFC-113/a (CFC-113 and CFC-113a collectively) and HCFC-142b (Fig. S2). Further, higher than expected inferred global emissions of substances used primarily as feedstocks suggest that the emitted fraction of production of these chemicals was higher than estimated during the 1990s^4,5,6,7,8, which underscores the potential benefits of limiting feedstock emissions, discussed for example by Andersen et al.⁹.

Evidence that feedstock-related emissions are higher than expected is exemplified here by CCl₄. Following the 2010 phase-out of CCl₄ for emissive uses, annual CCl₄ emissions remained at a level of several tens of thousands of tonnes between 2010 and 2023^7,10,11. These emissions could not be reconciled by the sum of (a) a 0.5% emission rate from its production and use as a feedstock, (b) additional emissions of CCl₄ as a by-product in the production of PVC (polyvinyl chloride), and (c) losses from landfills and other legacy sites^8,11,12. To close this gap, higher losses from feedstock CCl₄ were proposed by Liang et al.¹¹, and Sherry et al.¹², and finally the emission factor was revised to be 4.3% of production^8,12. In a recent analysis¹³, this value was further split into 2% losses during the actual production and 2.3% from the use as feedstock. It should be emphasised that this higher emission factor was independently accompanied by a substantial increase in the use of CCl₄ as a feedstock of 4–5% per year between 2014 and 2024 (Fig. S2). This development was mainly caused by a continuously increasing production of CCl₄-dependent HFOs and HCFOs (Fig. 1), which are important fluorochemical substitutes for enabling the phase-down of HFCs with larger GWPs. In addition to CCl₄, also emissions of CFC-113/a, CFC-114/a and 1,1,1-trichloroethane (CH₃CCl₃) have been higher than expected in recent years^{4,5,6,8,10,14,15}. Because emissions from remaining banks of these substances should now be negligible, their ongoing emissions must primarily be related to losses from continued feedstock use or from by-product emissions (see Fig. 1 for pathways).

For feedstock chemicals other than CCl₄, emissions were estimated by Daniel and Reimann et al.⁸ to be 2–4%, based on the ratio of global emissions, derived from atmospheric measurements, against production reported to UNEP³. Subsequently, this top-down estimate was further substantiated at 3.6%¹³ using bottom-up data, which is a combination of emissions from production (2.5%), distribution (0.5%), and the conversion of the ODS feedstocks into the final product (0.6%). This emission fraction of 3.6% is used in this work as the default for the business-as-usual (BAU) scenario for ODS feedstocks (except CCl₄) over the period 2024–2100. This same value is also used for Halon-1301, because we consider that there is no conclusive evidence for the substantially higher value of 26% that has been discussed¹³. For CCl₄, the slightly higher emission rate of 4.3%, discussed above, is applied⁸. Finally, the emission rate of CFC-115, likely only emitted as a by-product from the manufacturing of HFC-125 (Fig. 1), is estimated at 0.8% (see Supplementary Information).

Starting in 2024, the BAU emissions scenario is compared with a low-emission scenario (LOW) and a zero-emissions scenario (ZERO). The LOW scenario assumes an emission rate of 0.5% relative to the production from 2024 onward, consistent with the value originally assessed from well-managed facilities in the 1990s. It is used to assess the impact of immediate compliance with those anticipated feedstock-abatement levels in contrast to the BAU scenario. The ZERO scenario is identical to the LOW scenario but assumes zero feedstock-related emissions from 2024 onward.

To project future trends in the production and consumption of ODS feedstocks from 2025 to 2100, historical production trends between 2014 and 2024 (Fig. S2) are combined with expected changes in specific feedstock applications (see Supplementary Information). This results in compound-specific growth rates (Table 1) that are projected into the future in three steps. First, for the next 10 years (2025–2034), second, until 2050 and, finally, with stable production quantities between 2050 and 2100 for all feedstock applications. In brief, the production and consumption of ODS feedstock chemicals to produce controlled HFCs (Fig. 1) is projected to decline in the late 2020s, related to the phase-down under the Kigali Amendment. On the other hand, the production of HFOs, as important replacement compounds, is assumed to continue to increase at a rate similar to 2014–2024. These two opposing trends particularly influence the future production of CCl₄ and make the prediction of its long-term behaviour challenging. As a best guess, the current 4% annual increase in production is assumed to continue only until 2034 and is expected to remain stable afterwards. Other feedstock chemicals and by-products related to the controlled HFC production exclusively, such as CFC-114/a, CFC-115 and HCFC-124/a, are projected to decline according to the Kigali Amendment restrictions.

In recent years, the volume of feedstock chemicals used to produce halogenated polymers has been increasing. HCFC-142b is the feedstock chemical in the production of PVDF (polyvinylidene fluoride), which is increasingly used in electrical car batteries. Therefore, an increase of 6% per year is used here until 2034, followed by 4% per year until 2050. For CFC-113/a, the projection is more challenging. First, around a third of the CFC-113/a is used in the production of HFC-134a, which is projected to decline. Second, the demand for CFC-113/a to produce polymers—notably polychlorotrifluoroethylene (PCTFE) from CFC-113—and different specific chemicals is rising. Finally, they can be used to produce HFO-1336mzz, whose future demand is difficult to estimate. Therefore, the increase of 2% per year for CFC-113/a, used for polymers and other chemicals, is only maintained until 2034 and then kept stable.

Based on the above projections, the emissions of all feedstock chemicals and the related atmospheric mixing ratios have been calculated for 2024–2100 (Fig. S4). This forms the basis for deriving the mid-latitude effective equivalent stratospheric chlorine (EESC), a metric that can be used to compare the evolution of ODS relative to their 1980 level, which is regarded as a benchmark for stratospheric ozone recovery. In addition, projected mixing ratios are combined with compound-specific radiative forcings to estimate their effect on Earth’s radiative balance and, by extension, their potential influence on climate.

Impacts on stratospheric ozone and climate until 2100

Here, the effect on global emissions and mixing ratios caused by losses from ODS used as feedstock are discussed, using the BAU, LOW and ZERO scenarios. Figure 2 displays estimated emissions from global measurements networks from 2010 to 2023 (see “Methods”), together with projected emissions between 2024 and 2050. The emissions of the LOW and ZERO scenarios are applied in 2024 without any transitional period and therefore instantly diverge from the BAU scenario. For the future emissions, uncertainties are calculated based on a wider range of scenarios, as detailed in Table 1 of the “Methods” section.

Fig. 2: Emissions of ODS feedstock chemicals.

Historic and future emissions of ODS (ozone-depleting substances) used as feedstocks (2010–2050) in Gg/year (A), Gg-CFC-11 eq./year (B) and in Tg CO₂ eq./year (C). Total emissions arise from feedstock production and usage and other sources, such as losses from banks and other legacy uses. Emissions until 2023 were derived from NOAA (National Oceanic and Atmospheric Administration) and AGAGE (Advanced Global Atmospheric Gases Experiment) measurements (see “Methods”). Emissions between 2024 and 2050 are shown for the business-as-usual scenario (BAU), with the black line representing total emissions from all sources. The red line shows emissions which would result for the low scenario (LOW), and the yellow line represents the ZERO scenario (projected emissions from banks and other legacies alone, without emissions from feedstock production and usage). Emissions resulting from the uncertainty ranges given in Table 1 are shown as dashed lines.

Both the BAU and LOW scenarios show a decline in ODS mass emissions (Gg/year) between 2024 and 2050 (Fig. 2A). This decrease occurs because rising emissions from feedstock chemicals are offset by declining emissions from HCFC banks and anticipated reductions from legacy CCl₄ sources. In the LOW scenario, emissions are expected to decline steadily, slightly above the ZERO scenario (without emissions from feedstock production). In contrast, the BAU scenario only shows a decline of around 50%, and emissions are expected to stabilize around 2045. This levelling of emissions is mainly driven by the on-going use of CCl₄ in the production of HFOs and by the expected high consumption of HCFC-22 and HCFC-142b to produce halogenated polymers (Fig. S3). The impact of these sustained elevated emissions on global mixing ratios until 2100 is shown in Fig. S4. Notably, the mixing ratios of CCl₄, HCFC-22, and HCFC-142b, substances with substantial expected feedstock production, are projected to cease declining by 2100.

To assess the effect of ongoing losses from feedstocks and banks on stratospheric ozone, emissions were expressed as CFC-11-equivalents (Fig. 2B). In both the LOW and ZERO scenarios, CFC-11 eq. emissions decline by 2050, from 90 Gg CFC-11 eq. in 2024 to 30 Gg CFC-11 eq. and 15 Gg CFC-11-eq, respectively. In contrast, under the BAU scenario, emissions in 2050 decline only to 63 (55–75) Gg CFC-11 eq. (Fig. 2B).

The gap between the BAU and the LOW scenarios is primarily caused by CCl₄, with additional substantial contributions from Halon-1301, CFC-113/a, HCFC-22, and HCFC-142b (Fig S3). If the emission factor for Halon-1301 were as high as 26%, instead of the assumed 3.6%¹³, it is worth noting that total projected BAU emissions in 2050 would increase by an additional 11 Gg CFC-11-eq.

To evaluate the impact of feedstock-related emissions on stratospheric ozone depletion, their contribution to mid-latitude EESC was calculated through 2100^8,16 (see “Methods”). Figure 3A compares the scenarios with 1071 ppt, which corresponds to the mid-latitude EESC level in 1980. This level is widely regarded as a benchmark for the stratospheric ozone recovery. In the ZERO and LOW scenario, the mid-latitude EESC returns to the 1980 value in 2065 and 2066, respectively. In the BAU scenario, this recovery is delayed by about 7 years, i.e. in 2073 (uncertainty range: 2072–2077). The most effective way to reduce this delay would be to lower feedstock emissions of CCl₄ and to a lesser extent CFC-113/a (Fig. S3), related to their elevated projected emissions compared to other feedstocks, their long atmospheric lifetime, and their strong impact on ozone depletion.

Fig. 3: Effect of feedstock emissions on stratospheric ozone and climate.

A Mid-latitude effective equivalent stratospheric chlorine (EESC) in ppt for individual ODS (ozone-depleting substances) used as feedstocks, including non-feedstock related contributions  (e.g. emissions from banks). ODS not largely used as feedstocks (CFC-11, CFC-12, CH₃Cl, CH₃Br and several halons) are grouped as Non-Feedstocks. The black and red lines show the total effect of the business-as-usual (BAU) and the low emission (LOW) scenario, respectively; with uncertainties as dashed black lines. The ZERO scenario (projected emissions from banks and other legacies alone, without emissions from feedstock production and usage) is shown as a yellow dashed line. Additionally, the mid-latitude EESC value of 1980 is shown as a red dashed line, as a marker for the return date for the mid-latitude stratospheric ozone. B Radiative forcing in W m^–2 as a measure for the influence of Non-Feedstocks and individual feedstock ODS emissions on climate within the BAU, LOW, and ZERO scenario. Lines as in (A).

Finally, the projected direct climate impact of feedstock emissions until 2050, expressed in CO₂-equivalents (CO₂-eq., 100-year time horizon), is shown in Fig. 2C. In the LOW scenario, feedstock-related CO₂-eq. emissions gradually decline until 2050, with non-feedstock emissions (ZERO scenario) from banks and other sources still accounting for about two-thirds of the total at that time. In the BAU scenario, emissions are expected to decline until around 2045, after which they stabilize. The forecasted difference of around 300 Tg CO₂-eq. between the LOW and the BAU scenarios in 2050 corresponds to around 0.8% of global anthropogenic CO₂ emissions in 2024¹⁷. The largest contributors to this difference are HCFC-22 and CCl₄, followed by HCFC-142b and CFC-113/a (Fig. S3).

Figure 3B illustrates the contribution of ODS emissions from non-feedstock and feedstock uses to direct radiative forcing until 2100, as a measure of their climate impact. In 2100, the difference in radiative forcing between the BAU and LOW scenarios is 28 (14–45) mW m^-2. The impact of the BAU feedstock consumption scenario compared to the LOW scenario is around 8% of the radiative forcing of all ODS in 2020¹⁰ or 5 times higher than the radiative forcing of the very potent greenhouse gas SF₆ (sulphur hexafluoride) in 2020¹⁰.

In summary, ODS used as feedstocks were excluded from the Montreal Protocol controls at a time when remaining and future emissions from this practice were thought to be too small to substantially affect the timing of the ozone layer recovery. If release rates of ODS from feedstock production were 0.5% or lower, consistent with early industry-based estimates, their current and future influence on stratospheric ozone would be very modest. As a co-benefit, this would have further strengthened the Montreal Protocol’s contribution to reducing associated direct radiative forcing.

However, current emissions are higher than anticipated and may further increase, substantially delaying the recovery of stratospheric ozone. Under the BAU scenario, this delay is projected to be 7 (6–11) years relative to the LOW scenario. The BAU scenario combines the projected increases in feedstock use and production related to current legislation (e.g. Kigali Amendment) with our best estimate of future end-product needs and potential environmental restrictions. For most feedstock chemicals, their trends in emissions in 2014–2024 were progressed only up to 2034 (i.e. 10 years). For some compounds, however, with higher potential for future growth (e.g. HCFC-22, HCFC-142b), increases were maintained until 2050. If all ODS feedstock production and related emissions will progress up to 2050 at the same rate as in the last decade, this would lead to an average increase in the return date for the stratospheric ozone in mid-latitude which is well above 10 years, instead of the 7 years estimated here. In addition, if the relative emissions of H-1301 feedstock were 26% rather than 3.6% (as suggested¹³), the return of the mid-latitude EESC to 1980 levels would be further delayed by an additional 4 years.

Methods

Derived historic measurement-based emissions until 2023

For the calculation of global measurement-based emissions, data from AGAGE (Advanced Global Atmospheric Gases Experiment) and NOAA (National Oceanic and Atmospheric Administration, USA) were used^7,18,19, applying a global 12-box model²⁰. Briefly, the transport and process model is a zonally averaged model of the atmosphere, divided at the equator and 30°N and 30°S, and the surface, 500 hPa and 200 hPa. Loss process uses a first order offline chemistry scheme. Emissions estimates use monthly mean measurements of background air from seven sites in the AGAGE network, which are representative of the meridional semi-hemispheres. For the NOAA network, globally averaged annual mole fractions represent weighted annual averages from measured monthly mole fractions made at remote locations. Data from these measurement sites were used Alert, Nunavut, Canada, 82.5° N 62.5° W (NOAA); Zeppelin Mountain, Svalbard, 78.9° N, 11.9° E (AGAGE); Summit, Greenland, 72.6° N 38.4° W (NOAA); Barrow, Alaska, USA, 71.3° N 156.6° W (NOAA); Mace Head, Ireland, 53.3° N 9.9° W (AGAGE, NOAA); Jungfraujoch, Switzerland, 46.6° N, 8.0° E (AGAGE); Niwot Ridge, Colorado, USA, 40.1° N 105.6° W (NOAA); Trinidad Head, California, USA, 41.0° N 124.1° W (AGAGE, NOAA); Mauna Loa, Hawaii, USA, 19.5° N 155.6° W (NOAA); Cape Kumukahi, Hawaii, USA, 19.6° N 154.9° W (NOAA); Ragged Point, Barbados, 13.2° N, 59.4° W (AGAGE); Cape Matatula, American Samoa, 14.2° S 170.6° W (AGAGE, NOAA); Kennaook / Cape Grim, Tasmania, Australia, 40.7° S 144.7° E (AGAGE, NOAA); Palmer Station, Antarctica, 64.8° S 64.1° W (NOAA); and South Pole, Antarctica, 90.0° S (NOAA).

Emissions are estimated using Bayesian inference, which includes errors and uncertainties due to transport, measurements and atmospheric lifetimes.

Calculation of the mid-latitude EESC

For each compound and for each scenario described in this paper, future surface mixing ratios are calculated from emissions from the BAU, LOW, and ZERO scenarios together with atmospheric lifetimes²¹ using a global box model²². These mixing ratios are then used to calculate the future EESC representative of mid-latitude ozone loss, using the formalism developed in Engel et al.²³. Furthermore, mixing ratios are used to estimate the radiative forcing by multiplying the mixing ratios by the radiative efficiencies²¹.

Projected feedstock production and consumption between 2025 and 2100 for the BAU and LOW emission scenarios

The relative feedstock emissions from the business-as-usual (BAU) and the low (LOW) emission scenarios are explained in this section and summarized in Table 1. The projected increase in the use of individual ODS feedstocks between 2025 and 2034 is related to their trends in the production and consumption between 2014 and 2024 (Fig. S1), taking into consideration some additional information, explained in Table 1. For CCl₄, an annual growth rate of 4% (2–6%) was used from 2025 to 2034, based on the expected high growth in the production of HFOs, as the main fluorinated replacement for the restricted HFCs. Due to the uncertainty connected with the future production of HFOs and the usage of CCl₄ as a feedstock for phased-down HFCs, production was then assumed to be stable afterwards. For HCFC-142b, an annual growth of 6% (4–8%) is justified by its main usage as a feedstock for polyvinylidene fluoride (PVDF), which is used in advanced electric car batteries (see below). Between 2035 and 2050, a smaller annual increase of 4% (2–6%), representing the longer-term growth perspective, was used. For others, such as 1,1,1-trichloroethane and HCFC-141b, the production and consumption as feedstock have declined in recent years (Fig. S2, Table 1), which is projected forward on an individual basis.

Table 1 Emissions of ODS (ozone-depleting substances) and their future business-as-usual (BAU) and low (LOW) emission evolution, related to (i) their production and consumption as feedstock chemicals and (ii) other sources, such as emissions from the declining but continuing production for emissive uses of HCFCs, emissions as by-products, emissions from banks, and unexplained emissions for CCl₄

The most important categories of current emissions are summarized in Table 1 (as shown in Fig. 1), separated into (a) feedstock emissions, (b) by-product emissions, (c) bank, legacy and unexplained emissions, and (d) emissions from emissive uses (only for HCFCs).