Unexplained high and persistent methyl bromide emissions in China

Global production, consumption, and emissions of long-lived ozone-depleting substances (ODSs) have depleted the stratospheric ozone layer and contributed to warming. The Montreal Protocol regulates ODSs, including methyl bromide (CH3Br), which has a lifetime of ~0.8 years, an ozone depletion potential (ODP) of 0.6, and a 20-year GWP of 9. CH3Br is a major contributor to stratospheric bromine, which is highly efficient at depleting ozone. CH3Br has both anthropogenic (fumigation: quarantine and pre-shipment, QPS; and non-QPS; solvents; feedstocks) and natural sources (oceans, biomass burning, ecosystems, specific crops). Under the Protocol, non-Article 5 countries phased out CH3Br earlier; Article 5 countries (including China) froze production/consumption in 2002, reduced 20% by 2005, and eliminated by 2015, with QPS exemptions. China’s non-QPS and QPS usage peaked at 3.5 and 2.1 Gg yr−1 (2000 and 2005). Although non-QPS consumption was scheduled to phase out by 2015, China obtained critical use exemptions of ~0.1 Gg yr−1 for 2015–2018 (ginger), delaying full non-QPS phase-out to 2019. National CH3Br emissions in China over the last decade remain uncertain due to limited observational coverage (previously largely the Gosan site, South Korea) and limitations of interspecies correlation methods, plus incomplete bottom-up inventories omitting sectors such as feedstock use and several terrestrial sources. This study quantifies China’s CH3Br emissions (2011–2020) using a 10-site observational network and atmospheric inversion, and constructs an improved bottom-up inventory, revealing spatial hotspots and a substantial unexplained emission component, with implications for ozone recovery and Montreal Protocol compliance.

Prior work used limited regional observations (e.g., Jeju Island/Gosan) and interspecies ratio approaches to estimate Chinese CH3Br emissions, yielding highly uncertain and often single-year estimates. Earlier bottom-up inventories omitted important sectors such as industrial feedstock/solvent uses and terrestrial sources (rice paddies, salt marshes, mangroves, fungi), likely underestimating emissions. Recent studies have identified new potential sources (e.g., baking; copper-based pesticides) but their contributions appear negligible at national scales. This study addresses these gaps by deploying a broader Chinese observation network, applying Bayesian inversion, and expanding bottom-up sectoral coverage to include additional anthropogenic and terrestrial sources.

Observations and sampling: Atmospheric CH3Br mole fractions were measured at ten sites across China (AKD, HYN, JGJ, JSA, LAN, LFS, WLG, XFG, XGL, SDZ) from 2011 to 2020, located >20 km from industrial/dense population centers to capture regional signals. Weekly or daily flask sampling used membrane pumps and stainless-steel canisters, analyzed centrally with a Medusa GC-MS (Agilent 6890/5975B). At SDZ, in situ measurements also operated 2011–2020. Measurements were linked to the AGAGE SIO-2005 calibration scale. Stability tests showed no significant drift; precision was ~1% (flask) and 0.6% (in situ). Baselines were estimated as the lowest observation in a 90-day moving window (flask sites), with cross-checks against AGAGE background methods; for SDZ in situ, a percentile-based method (lowest 25% in 30-day windows minus simulated enhancements) was used. Backgrounds at four AGAGE sites (MHD, THD, GSN, RPB) were used for comparison and sensitivity tests. Emission sensitivity: FLEXPART (Lagrangian dispersion model) driven by ECMWF 3-hourly, 1°×1° meteorology performed 20-day backward simulations releasing 40k particles every 3 hours at each site, deriving source–receptor relationships (footprints). Chemical loss of CH3Br during 20-day back-trajectories was neglected given small expected loss (~1.7%) and minimal impact shown for similar species. Bayesian inversion: A Bayesian framework minimized a quadratic cost function to obtain posterior emissions on a gridded domain using observations minus baselines, footprints (H), and prior emissions with uncertainties (Sa, So). Prior national emissions were set to 5 Gg yr−1, spatially disaggregated by population distribution; sensitivity tests scaled priors by 0.5 and 1.5 and set prior uncertainties to 100%, 150%, and 200%. Nine inversions (3 priors × 3 uncertainties) formed an ensemble; the final posterior was the ensemble mean with posterior uncertainty from the ensemble and inversion covariance. Daily means were used for SDZ to reduce temporal correlation; observational error covariance was diagonal using annual site-specific standard deviations. Regions with low sensitivity were masked from totals. Bottom-up inventory: Emissions were estimated for four sectors: (1) Fumigation (QPS and non-QPS) using reported consumption (UNEP) and emission factors EFnon-QPS=65%, EFQPS=84%. (2) CH3Br production and non-fumigation uses: industrial production leakage and feedstock uses (EF=4% for feedstock and production), and solvents/cleaning agents assumed fully emitted over two years (50% per year). (3) Combustion: open biomass burning using GFAS dry matter burned, MODIS MCD12C1 land cover, biomass-specific emission factors; indoor biofuel burning from agricultural residues using production statistics, residue factors, burn ratios, dry matter fractions, combustion efficiency, and an emission factor of 0.0011 g kg−1. (4) Terrestrial ecosystems: rice paddies (area × flux 1.1 mg m−2), rapeseed (global emissions scaled by China’s planting area share), salt marshes and mangroves (global sector totals scaled by China’s area shares), and fungi (global fungi emissions scaled by China’s vegetation area share). Uncertainties: Monte Carlo sampling (100,000 draws) applied with 10% uncertainty on production/consumption data, 5% on agricultural and area statistics; sectoral emission factor uncertainties followed literature sources (see supplementary tables). Ozone depletion metrics: ODP-weighted emissions (CFC-11-eq) computed by multiplying mass emissions by ODP (0.6) and chlorine-equivalence where appropriate; Integrated Ozone Depletion (IOD) estimated for sustained missing emissions to 2050 using IOD = K × EEq × (τatmos/τstrat) with K=100±16 DU years per Tg Cl and a bromine efficiency factor of 60 for CH3Br.

• Measured CH3Br mole fractions at 10 Chinese sites (2011–2020) consistently exceeded background levels, with strongest enhancements in eastern China; site mean enhancement ranged from 0.6 ± 0.8 ppt (WLG) to 9.9 ± 1.4 ppt (LAN).
• Top-down national CH3Br emissions averaged 9.2 ± 1.4 Gg yr−1 over 2011–2020, with interannual variability but no significant trend (slope 0.16 ± 0.15 Gg yr−1, p=0.33); maximum 10.5 ± 1.4 Gg yr−1 (2014) and minimum 6.2 ± 1.1 Gg yr−1 (2012).
• Spatial distribution indicates a hotspot in eastern China (Anhui, Beijing, Hebei, Jiangsu, Liaoning, Shandong, Shanghai, Tianjin, Zhejiang, Fujian, Guangdong): flux 4.0 vs 0.53 kg km−2 yr−1 elsewhere, contributing ~52% (4.8 Gg yr−1) of national emissions from only 12.5% of the land area.
• Top five emitting provinces and mean emissions: Jiangsu 0.76 ± 0.16, Shandong 0.74 ± 0.12, Zhejiang 0.72 ± 0.23, Anhui 0.66 ± 0.19, Guangdong 0.50 ± 0.17 Gg yr−1 (~40% of national total).
• Bottom-up inventory averaged 3.7 ± 0.2 Gg yr−1 (2011–2020): fumigation (QPS + non-QPS) ~2.0 Gg yr−1 (54% of bottom-up). Non-QPS 0.07 Gg yr−1 (consistent with UNEP 0.06); QPS 1.96 Gg yr−1, exceeding reported 0.91 Gg yr−1, pointing to underreporting or underestimated emission factors/activity.
• Seven additional sectors beyond earlier inventories (industrial production, feedstock, solvents; rice paddies, salt marshes, mangroves, fungi) contributed 1.0 ± 0.1 Gg yr−1 (~26% of bottom-up), comparable to UNEP-based fumigation emissions.
• Large unexplained emissions: top-down minus bottom-up gap of 5.5 ± 1.4 Gg yr−1 (~60% of top-down), persistent through 2011–2020.
• Potential causes: (1) illegal production/sales events documented 2010–2014 and in 2020 (e.g., Shandong); (2) underestimation of terrestrial sources (temperature, soil bromide, plant type dependence; sparse flux data); (3) unidentified sources (recently discovered sources like baking and copper-based pesticides appear negligible in China).
• Ozone impact: ODP-weighted CH3Br emissions from China estimated at 5.5 ± 0.78 CFC-11-eq Gg yr−1; by 2019, CH3Br national emissions exceeded eastern China’s CFC-11 emissions. If China’s missing source persists to 2050, integrated ozone depletion is estimated at ~20 DU years; larger global impacts possible if similar unidentified sources exist elsewhere.

The study resolves China’s national CH3Br emissions and their spatial distribution using a dense observational network and inversion. Despite the phase-out of non-QPS uses (with limited CUEs through 2018), emissions remained persistent with no significant downward trend during 2011–2020, indicating substantial ongoing sources. The top-down estimates significantly exceed the comprehensive bottom-up inventory, revealing that known, inventoried sectors account for only about 40% of observed emissions. The disproportionate emissions in eastern provinces align with concentrated port activities and QPS-related fumigation, industrial uses, and population density, yet the magnitude still implies additional sources. The identified gap likely arises from a combination of underreported/illegal production and trade, underestimation of terrestrial emissions (due to sparse, variable flux data), and yet-unidentified sources. These findings underscore compliance and enforcement challenges under the Montreal Protocol and highlight that CH3Br’s relative importance among ODSs in China is growing as legacy ODS emissions decline. The observed discrepancies call for improved reporting, targeted source investigations in emission hotspots, enhanced flux measurements of terrestrial sources, and expansion of atmospheric monitoring to reduce uncertainties and reconcile top-down and bottom-up budgets.

This work delivers the first multi-year, China-wide top-down estimates of CH3Br emissions (2011–2020) constrained by a national monitoring network and a robust inversion framework, alongside an expanded bottom-up inventory incorporating additional industrial and terrestrial sectors. Emissions averaged 9.2 ± 1.4 Gg yr−1 with eastern China as a dominant hotspot, while the bottom-up total (3.7 ± 0.2 Gg yr−1) leaves a large unexplained gap (~5.5 ± 1.4 Gg yr−1). The persistence of high emissions after the scheduled phase-out points to underreported or illegal activities, underestimated natural/terrestrial sources, and potentially unidentified sources. Given the sizable ODP-weighted contribution and projected integrated ozone depletion from sustained missing emissions, urgent actions include: improving production/consumption data transparency and verification; conducting representative flux measurements for key terrestrial sources across climates and land uses; investigating industrial and logistics chains (especially QPS-related fumigation) in hotspots; and expanding observational coverage to better constrain regional emissions. These steps are essential to close the budget gap and safeguard progress toward ozone layer recovery under the Montreal Protocol.

• Observational coverage, while broad, leaves regions of low sensitivity that were excluded; additional sites would improve regional constraints.
• Baseline estimation and inversion rely on assumptions (e.g., neglecting chemical loss over 20-day back-trajectories) that introduce small but nonzero uncertainties.
• Prior emissions and uncertainties (population-based disaggregation; ensemble scaling) may not perfectly represent spatial patterns.
• Bottom-up estimates depend on activity data quality, emission factors, and scaling of global sector totals to China (e.g., salt marsh, mangrove, fungi), which may not capture local variability.
• Sparse and heterogeneous terrestrial flux measurements limit accuracy in natural/biogenic source estimates.
• Potential underreporting in QPS consumption and uncertainties in industrial leakage factors and solvent emission assumptions affect anthropogenic sector estimates.
• Some newly identified sources (e.g., copper-based pesticide-related emissions) were approximated using surrogate data (total CuSO4 consumption), adding uncertainty.