Increase in global emissions of HFC-23 despite near-total expected reductions

Hydrofluorocarbons (HFCs) are potent greenhouse gases and are regulated under the Kyoto Protocol and the Kigali Amendment to the Montreal Protocol to prevent offsetting the climate benefits of CFC and HCFC phaseouts. HFC-23 has a very long atmospheric lifetime (~228–240 years) and a high radiative efficiency, yielding the highest 100-year global warming potential among HFCs (GWP100 ≈ 12,690). HFC-23 is mainly generated as an unwanted by-product during production of HCFC-22, with smaller emissions from feedstock use, semiconductor manufacturing, low-temperature refrigeration, and specialty fire suppression. Previous atmospheric observations and firn air measurements showed HFC-23 mole fractions rising from near-zero in 1960 to 28.9 ± 0.6 pmol mol−1 by end-2016. Top-down studies inferred global emissions rising from 4.2 ± 0.7 Gg yr−1 (1980) to 13.3 ± 0.8 Gg yr−1 (2006), declining to 9.6 ± 0.6 Gg yr−1 (2009) during Clean Development Mechanism (CDM) abatement, then increasing to 14.5 ± 0.6 Gg yr−1 (2014) and declining to 12.7 ± 0.6 Gg yr−1 (2016). This study aims to update global HFC-23 emissions through 2018 using atmospheric observations and to compile a bottom-up inventory through 2017 incorporating reported post-2015 abatement by China and India, to assess whether expected reductions occurred and to quantify any discrepancies.

The paper synthesizes prior top-down and bottom-up estimates of HFC-23. Earlier work linked HFC-23 emissions to HCFC-22 production and showed the effect of CDM-driven destruction in reducing emissions around 2006–2009. Studies reported resumed growth post-CDM as abatement waned. The literature establishes methods: AGAGE measurements, firn air, box-model inversions, and inventory-based estimates using UNFCCC and UNEP data, as well as CDM monitoring reports. Prior inference suggested that reductions during 2014–2016 might align with Chinese actions, but the magnitude and persistence of reductions were uncertain.

• Atmospheric observations: In situ HFC-23 measurements from five core AGAGE background stations (Mace Head, Trinidad Head, Ragged Point, Cape Matatula, Cape Grim), using Medusa GC-MS on the SIO-07 calibration scale, up to 12 calibrated measurements per day. Pre-2007 data from the Cape Grim Air Archive and Northern Hemisphere flasks analyzed with the same system. Precision ~0.5–1% for daily reference gas measurements; system blanks and lab air checks weekly.
• Top-down emissions inference: AGAGE 12-box 2-D atmospheric chemistry-transport model (four equal-mass latitude bands, three vertical levels at 1000/500/200 hPa). HFC-23 lifetime ~240 years (hydroxyl reaction and stratospheric loss). Bayesian inversion with weak a priori constraint on emissions growth rate (prior growth rate 0 ± 20% of maximum bottom-up emissions), measurement uncertainties based on monthly baseline variability for in situ data and combined terms for archive data; posterior uncertainties include calibration (~3%) and loss frequency (~21%). Meteorology is annually repeating; uncertainty from this could not be assessed.
• Bottom-up inventory: Developed-country emissions from 2019 UNFCCC National Inventory Reports. Developing-country emissions estimated from UNEP HCFC-22 production data (dispersive and feedstock combined) multiplied by time-varying emission factors (historically ~4%, declining to ~3%; post-2012 factors from Multilateral Fund reports). Two scenarios: (i) developing countries no abatement; (ii) developing countries with abatement, incorporating CDM destruction (2003–2014; compiled from 474 monitoring reports across 19 facilities) and post-2015 reported abatement (China’s HPPMP: 45%, 93%, 98% reductions in 2015–2017; India’s 13 Oct 2016 executive order mandating destruction). Global totals combine developed and developing estimates.
• Comparisons: Forward model runs using bottom-up with abatement assessed expected mole fraction growth; discrepancies between top-down and bottom-up quantified annually. Integrated differences 2015–2017 converted to CO2-equivalent using HFC-23 GWP.

• Contrary to expectations of an 87% drop in global HFC-23 emissions between 2014 and 2017 due to Chinese and Indian abatement, atmospheric observations show increases, reaching a record 15.9 ± 0.9 Gg yr−1 in 2018.
• AGAGE-derived global mean mole fraction growth rate increased from 2016, reaching 1.1 ± 0.05 pmol mol−1 yr−1 in 2018 (global mean mole fraction 31.1 pmol mol−1).
• Bottom-up with abatement estimates fell to 2.4 ± 0.9 Gg yr−1 in 2017, the lowest in 17 years, but disagreed with top-down by 12.5 ± 0.7 Gg yr−1.
• Reported post-2015 reductions: China’s HPPMP estimated reductions of 6.1, 13.0, and 15.2 Gg yr−1 in 2015, 2016, 2017; India’s reductions of 0, 0.4, and 1.9 Gg yr−1. Combined, 6.1, 13.4, and 17.1 Gg yr−1 expected.
• Observed decline from 2014 to 2016 in top-down emissions was only ~1.6 ± 1.6 Gg (2σ) (~11%), far smaller than anticipated from reported Chinese reductions (~13 Gg), and coincided with a modest 6.3% drop in HCFC-22 production.
• In 2014, bottom-up no-abatement and with-abatement estimates converged at ~20.8 Gg yr−1 as CDM abatement ended, while top-down was 14.5 Gg yr−1. If maximum CDM abatement (9.0 Gg yr−1 from 2011) had continued, bottom-up would be ~11.8 Gg yr−1 in 2014.
• Integrated discrepancy (top-down minus bottom-up with reported abatement) was 24.4 Gg over 2015–2017, corresponding to ~309 Tg CO2-equivalent additional emissions.
• Reconciling reported Chinese reductions with observations would require implausible increases: ~780% rise in developed-country emissions, ~690% rise from India, or ~4250% unreported HCFC-22 production (unabated) relative to reported 2017 totals.

The study compares observation-inferred (top-down) emissions with inventory-based (bottom-up) estimates that incorporate reported abatement measures. The persistent and increasing top-down emissions through 2018, despite reported post-2015 abatement, indicate that the anticipated large reductions were not realized at the global scale by 2017–2018. The modest decline from 2014 to 2016 likely reflects short-term reductions in HCFC-22 production rather than effective sustained abatement. The magnitude of the mismatch suggests either that reported abatement—particularly in China, the dominant producer—was not fully implemented or effective, or that substantial unreported HCFC-22 production occurred with venting of HFC-23 by-product. Given the implausibility of the alternative explanations (very large increases from developed countries or India alone), the results underscore gaps in implementation or reporting. These findings are highly relevant to climate mitigation, as missed HFC-23 abatement represents large CO2-equivalent emissions. The work demonstrates the critical role of atmospheric observations in independently verifying emission inventories and policy outcomes.

Global HFC-23 emissions reached a historical maximum in 2018, contrary to expectations of near-total reductions from reported abatement in major producing countries. The large discrepancy between top-down and bottom-up estimates for 2015–2017 implies that pledged reductions were not fully realized and/or that unreported HCFC-22 production led to significant unaccounted HFC-23 emissions. Addressing this gap could have avoided an estimated ~309 Tg CO2-equivalent emissions over 2015–2017. Rapid, verifiable implementation of abatement at HCFC-22 facilities, improved transparency and reporting for HCFC-22 production (including feedstock uses), and continued atmospheric monitoring and inverse modeling are recommended. Future work should refine regional source attribution, enhance verification frameworks, and assess facility-level performance of HFC-23 destruction systems.

• The inverse modeling uses annually repeating meteorology; the uncertainty introduced by this assumption could not be assessed.
• Posterior uncertainties depend on assumed prior and measurement errors; systematic uncertainties in calibration (~3%) and loss frequency (~21%) remain.
• Developing-country HCFC-22 production data are aggregated (country-level data confidential), limiting detailed attribution.
• The study infers emissions globally; it cannot directly verify facility-level abatement performance or detect specific unreported production sites.
• Some discrepancies may stem from uncertainties in emission factors for bottom-up estimates and potential temporal variability in HCFC-22 production not captured by reporting intervals.