Top-down and bottom-up estimates of anthropogenic methyl bromide emissions from eastern China

. Methyl bromide (CH 3 Br) is a potent ozone-depleting substance (ODS) that has both natural and anthropogenic sources. CH 3 Br has been used mainly for preplant soil fumigation, post-harvest grain and timber fumigation, and structural fumigation. Most non-quarantine/pre-shipment (non-QPS) uses have been phased-out in 2005 for non-Article 5 (developed) countries and in 2015 for Article 5 (developing) countries under the Montreal Protocol on Substances that Deplete the Ozone 20 Layer; some uses have continued under critical use exemptions (CUEs). Under the Protocol, individual nations are required to report annual data on CH 3 Br production and consumption for quarantine/pre-shipment (QPS) uses, non-QPS uses, and CUEs to the United Nations Environment Programme (UNEP). In this study, we analyzed high precision, in situ measurements of atmospheric mole fractions of CH 3 Br obtained at the Gosan station on Jeju island, Korea, from 2008 to 2019. The background mole fractions of CH 3 Br in the atmosphere at Gosan declined from 8.5 ± 0.8 ppt in 2008 to 7.4 ± 0.6 ppt in 2019 at a rate of - 25 0.13 ± 0.02 ppt yr -1 . At Gosan, we also observed periods of persistent mole fractions (pollution events) elevated above the decreasing background in continental air masses from China. Statistical back trajectory analyses showed that these pollution events predominantly trace back to CH 3 Br emissions from eastern China. Using an inter-species correlation (ISC) method with the reference trace species CFC-11 (CCl 3 F), we estimate anthropogenic CH 3 Br emissions from eastern China at an average of 4.1 ± 1.3 Gg yr -1 in 2008–2019, approximately 2.9 ± 1.3 Gg yr -1 higher than the bottom-up emission estimates reported to 30 UNEP. Possible non-fumigation CH 3 Br sources - rapeseed production and biomass burning – were assessed and it was found that the discrepancy is more likely due to unreported or incorrectly reported QPS and non-QPS fumigation uses. These unreported anthropogenic emissions of CH 3 Br are confined to eastern China and account for 30–40% of anthropogenic global

3 not controlled (exempted from phase-out) under the MP have remained relatively constant over the past 20 years, and now account for more than 98% of the estimated consumption of CH3Br currently reported due to the phase-out of other regulated uses (TEAP, 2020). Despite no formal regulation, most parties to the MP are making efforts to minimize the use of CH3Br for QPS use and replace it with suitable alternatives such as heat treatment, phosphine (PH3), ethyl formate (C2H5OCHO), sulfuryl fluoride (SO2F2) and ethanedinitrile (NCCN). As a consequence of this CH3Br phase-out, the global atmospheric mole fraction 70 of CH3Br decreased from 9.2 parts per trillion (ppt) at the peak in 1996-1998, to 6.6 ppt in 2015, but then showed a slight positive growth of 0.14 ppt yr -1 (2.1% yr -1 ) from 2015 to 2016 (Engel and Rigby et al., 2019).
Global anthropogenic emissions of CH3Br can be estimated using "bottom-up methods from consumption and production data across various activities reported to UNEP annually by individual nations using activity-dependent emission factors (e.g., 65% for reported non-QPS consumption and 84% for the reported QPS consumption; MBTOC, 2006). Significant uncertainties 75 result from the emission factors and the speciation of CH3Br consumption across various activities (Vaughn et al., 2018). As QPS uses of CH3Br are generally highly emissive, consumption for these activities can be more accurately converted into emissions for this application (MBTOC, 2018).
"Top-down" estimates of global CH3Br emissions are derived from modelling of measured atmospheric mole fractions and atmospheric transport processes, for example using the AGAGE 12-box model of the atmosphere, assuming an atmospheric 80 lifetime for CH3Br (Cunnold et al., 1994;Prinn et al., 2005;Rigby et al., 2013). Regional characteristics of CH3Br emissions however cannot be obtained with the AGAGE 12-box or similar model, because they do not have the resolution to account for the synoptic scale of the atmospheric flow. Since the MP control of CH3Br consumption applies at a national level, rather than globally, it is important to estimate the top-down emissions at a regional to national scale (Weiss and Prinn, 2011). China is the largest producer and consumer of agricultural products in the world and therefore has potentially large anthropogenic 85 sources of CH3Br and is an important region for understanding CH3Br emissions in East Asia. Several studies have estimated the regional or national emissions from China based on "top-down" approaches using atmospheric observations. Blake et al. (2003) estimated the CH3Br emissions of 2.6 Gg yr -1 in China (South China: 2.0 Gg yr -1 and North China: 0.6 Gg yr -1 ) from aircraft observations in 2001. An inverse modeling study (Vollmer et al., 2009) using high-frequency ground measurements suggested emissions from China had decreased to 0.24 Gg yr -1 in 2006-2008. However, those results were based on a limited 90 period of observations (e.g., few months to years) and could not analyze the long-term variations and trends in CH3Br emissions.
Since then, there have been no further studies tracking the CH3Br emission trends in East Asia.
In this study, we present the 12-year high-precision, high-frequency record of atmospheric CH3Br mole fractions observed at Gosan station on Jeju island, South Korea, and analyze the observed variations in atmospheric CH3Br. We estimate annual emissions of CH3Br mainly from anthropogenic sources in eastern China, based on the empirical inter-species correlations 95 between CH3Br and CFC-11 during pollution episodes from eastern China and the well-defined eastern  emissions. This is the first study to present the long-term changes in CH3Br emissions from eastern China, after the phase-out period. In the following sections, in Section 2, we first introduce the Gosan station and the in situ ground-based instrumentation for CH3Br measurements, and long-term seasonal and annual variations of atmospheric CH3Br mole fractions are discussed. mass back-trajectory statistics, and describe the interspecies correlation method to estimate emission of CH3Br. In Section 3 and 4, the observation-based emission estimates of CH3Br in eastern China are further discussed considering the existing discrepancy between the global bottom-up and top-down emissions of CH3Br.

Instrumentation and measurement data 105
The coastal atmospheric observation station Gosan (GSN, 33.3°N, 126.2°E, 72 m a.s.l) at the south-western tip of Jeju island, South Korea (See Figure 1) is ideally located to monitor regional background mole fractions of atmospheric trace gases due to minimal influence of local anthropogenic pollution sources, and the strong pollution outflows from China, Korea, and Japan in East Asia (Kim et al.2012;Li et al., 2011 and. The in situ measurement system at Gosan, a "Medusa" gas chromatography-mass spectrometer (GC-MS) equipped with a 110 cryogenic pre-concentration system (Miller et al., 2008;Prinn et al., 2018), monitors more than 40 halogenated compounds including CFC-11 and CH3Br. As a part of the Advanced Global Atmospheric Gases Experiment (AGAGE; Prinn et al., 2018), Gosan station has been conducting continuous high-precision and high-frequency observations approximately every 2-hours (12 times per day) from 2008 to the present. The precisions (1 ) of all species, determined from repeated analysis (n=12) of an ambient standard, are better than 1% (i.e., the precision of CH3Br < 0.1%). The atmospheric abundances of most of the 115 Medusa compounds are calibrated on scales maintained by the Scripps Institution of Oceanography (SIO) (e.g., SIO-05 scale for CH3Br in this study).
Long-term, high-frequency CH3Br data observed during 2008-2019 at Gosan and background mole fraction data from Mace Head, Ireland (53. 3°N, 9.9°W) and Cape Grim, Australia (40.7°S, 144.7°E) are shown in Figure 2. Mace Head and Cape Grim, the primary sites of AGAGE, have been measuring various well-established trace gases including halogen compounds in the 120 atmosphere, for a long time and are historically representative remote background monitoring stations for the Northern and Southern Hemispheres, respectively (Prinn et al., 2018). Therefore, they are suitable sites to evaluate the measurement performance and seasonal variation of CH3Br at Gosan.
Regional background mole fractions of CH3Br were determined by removing pollution events after applying a polynomial fit to the lower 99.7% (within 3 ) of the Gaussian distribution derived from the 121-day observations for 60 days before and after 125 each observed data point . The baseline mole fraction at Gosan and Mace Head (northern hemisphere) are higher than those of Cape Grim (southern hemisphere), while the annual cycles at Gosan and Mace Head are similar.
The annual average CH3Br baseline mole fraction at Gosan decreased steadily from 8.5 ± 0.8 ppt in 2008 to 7.4 ± 0.6 ppt in 2019 (Table 1), declining at a rate of -0.13 ± 0.02 ppt yr -1 (-1.5% yr -1 ). This rate of decline for CH3Br is consistent with the global trend of atmospheric CH3Br determined from AGAGE in situ and NOAA (National Oceanic and Atmospheric 5 Administration) flask data in 2011-2012 that reported period in Carpenter and Reimann et al., 2014, which has been attributed to the influence of the CH3Br restrictions on non-QPS use imposed by the Montreal Protocol.
The monthly mean CH3Br baseline mole fractions for 2008-2019 are shown in Figure 3. The seasonal variations show a steady increase in spring, reaching a maximum in May, then dropping in June-July, followed by a constant level for the last 5 months of the year. The various sources and sinks of CH3Br likely show seasonal variability and the summertime minima in CH3Br 135 can be largely explained by the atmospheric mole fractions of OH reaching a maximum during the boreal summer (Cox, 2002;Simmonds et al., 2004) and long-range transport of southern hemispheric air parcel that over-cross the tropical regions .
Despite the continuous decrease in background mole fractions, we observed clear pollution signals (shown in red in Figure 2) through the entire study period, representing persistent inflow to Gosan of air masses influenced by regional CH3Br emission 140 sources and thus containing elevated mole fractions of CH3Br. The annual means of the enhancement mole fraction (pollution baseline; hereafter, enhancement) are consistently in a range of 3.6 to 4.6 ppt as given in

Statistical method to identify the potential CH3Br source regions
The regional distribution of potential CH3Br sources in East Asia was derived by applying statistical analysis of back trajectories corresponding to the observed CH3Br enhancements at Gosan from 2008 to 2019. Air mass back trajectories were generated using the Hybrid Single-Particle Lagrangian Integrated Trajectory (HYSPLIT; Stein et al., 2015) model from the NOAA Air Resources Laboratory with meteorological data output from the Global Data Assimilation System (1˚×1˚ horizontal 150 resolution, 23 vertical layers, ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1). The HYSPLIT 6-day air mass backward trajectories were initialized 500 m above the Gosan observation station, at a height where topographical influences can be minimized . To minimize the error that arises from a small number of outlier trajectories, only grids with more than 12 over-passing trajectories were used to define a potential source region ; method described in SI). China and southern Korea. In particular, high potential source regions for CH3Br emissions are seen along the Yangtze River that connects Shanghai, Nanjing, Hefei and Wuhan. The port of Shanghai is one of the busiest container ports in the world since 2010, with high volumes of port traffic and a large population (Robert et al., 2020). For example, Shanghai handled 43.3 million twenty-foot equivalent units in 2019 (https://safety4sea.com/port-of-shanghai-worlds-busiest-container-port-for-2019/, last access: 11 March, 2021). In several cases, high CH3Br mole fractions were observed in narrow-width air mass back 160 trajectories that showed long residence times over the port of Shanghai (see Figure S1; that additionally simulated by FLEXPART to confirm the dispersion effect instead of single trajectories of HYSPLIT), which would be consistent with Shanghai being a likely major port for QPS usage of CH3Br.
The high potential source regions include not only modern industrial urban areas but also the vast alluvial plains along the Yangtze River and its main tributaries. Note that this statistical analysis has little sensitivity to emissions from southwestern-165 western China and tends to over-estimate source strengths near the modeling boundary due to the limits of the 5-6 day backward trajectory domain of the HYSPLIT model. Therefore, those parts of China have been excluded from further discussion . Also note that this statistical trajectory analysis tends to underestimate emissions at sub-grid scale hotspots because the measured concentration gets distributed evenly over the grid cell (Stohl, 1996). Also, the dilution effects on distant source emissions are not considered in this statistical approach (Vollmer et al., 2006). Thereby, the emissions 170 from nearby sources might be overestimated due to the higher CH3Br concentration. For this reason, the emission potential for South Korea, shown in Figure 5, maybe lower. We do not attempt to identify more exact locations of CH3Br emission sources based on this approach because of its potential uncertainties, nevertheless it is clear that significant emissions of CH3Br originate predominantly from eastern China and South Korea.

Interspecies correlation method to estimate emission of CH3Br 175
In the previous section, it was noted that most of the air masses exhibiting enhanced CH3Br mole fractions flow into Gosan from China and Korea. We classify the air mass origins into 17 regions (see Figure S2a for the regional domains) based on the 6-day kinematic back trajectories of the HYSPLIT model. If a trajectory arriving at Gosan had entered the boundary layer (as defined by HYSPLIT) only within the regional domains for eastern China-1 (region 15), eastern China-2 (region 16), and Shandong provinces (region 17), it was defined as an air mass originating from eastern China. The air mass classification 180 applied to the CH3Br time series is illustrated in Figure S2b. The proportions of CH3Br pollution events from 2008 to 2019 classified into China, North and South Korea, and other regions were 37, 44, and 19%, respectively. Among them, 98% of air masses classified as China correspond to eastern China (~35% of the total).

Interspecies correlation method 185
To estimate emission of CH3Br from eastern China, we applied an interspecies correlation (ISC) method (Palmer et al., 2003;Dunse et al., 2005;Yokouchi et al., 2006;Millet et al., 2009;Li et al., 2011;Shao et al., 2011;Wang et al., 2014;. Described as a "ratio-method", this approach can derive the emission of a trace gas of interest from the correlation of its enhancement above baseline with that of a reference compound. This empirical ratio approach can estimate regional emissions of various substances in a simple and robust manner, compared to inverse methods that require complex 190 computational processes in combination with chemical transport models. For a reference tracer in ISC method, the following conditions are required: i) long lifetime, thus low chemical reactivity during transport from source to observation site, ii) wellquantified emission sources, iii) approximate co-location of the source regions for the reference and target species resulting in significant correlations with the target species. Several previous studies have used carbon monoxide (CO) as a reference species for ISC (Palmer et al, 2003;Dunse et al., 2005;Guo et al., 2009;Wang et al., 2014). CO can be observed readily at the target 195 species observation sites and often has documented emissions from anthropogenic sources, usually as a component of regional air quality emissions inventories. One of the issues of using CO as a reference species in ISC is that the emission inventories usually document anthropogenic CO sources only, whereas the observations see CO emissions from anthropogenic and natural sources, biomass burning for example. Therefore, in the CO observational data record, CO pollution episodes have to be identified as predominantly anthropogenic before inclusion in the ISC emissions calculations, which complicates matters. 200 Instead, we selected CFC-11 as the reference compound, because CFC-11 has a long lifetime (50-60 years) with low chemical reactivity, has been measured simultaneously with CH3Br at Gosan showing strong correlations (Li et al., 2011), and the CFC- The emissions of CFC-11 are from anthropogenic sources onlythere are no natural sources of CFC-11. Although the emission sources of CFC-11 and CH3Br are not necessarily co-located on an emission activity basis, we can still apply ISC method to 210 estimate the magnitude of country/regional scale emissions of CH3Br, when they occur within a same country/region where CFC-11 is emitted. When the likely CFC-11 and CH3Br sources are not co-located on a fine scale but are co-located on a regional scale, then it is important to make the CFC-11 and CH3Br observations sufficiently distant (hundreds of km) from the source region so that the initial individual plumes of CFC-11 and CH3Br emissions from separate sources become well mixed.
In this study, the emissions of CH3Br in eastern China are derived using the following equation: 215 where, and −11 are the emissions of CH 3 Br and CFC-11, respectively, is a slope of the linear regression between enhancements of CH3Br and CFC-11 (∆CH3Br and ∆CFC-11), and −11 are the molecular weights of CH3Br and 220 CFC-11, respectively. The intercept term of the linear regression can be ignored because it is generally not significantly different than zero, confirmed by the similar slope terms from linear and linear-through-the-origin regressions (Dunse et al., 2005).
The uncertainty of CH3Br emissions is associated with uncertainties of and −11 and determined by an error propagation method as follows: 225 where, is the uncertainty of estimated CH3Br emissions, −11 and are the uncertainties of −11 and , respectively. 230

Emissions of reference tracer
We use known emission estimates of CFC-11 from eastern China, which were derived by inverse modelling of Gosan CFC-11 observation data (Rigby et al., 2019;. Atmospheric mole fractions for CFC-11 observed over the same period with CH3Br are shown in Figure S3. CFC-11 emissions were estimated from four different Bayesian inverse methods based on two different Lagrangian atmospheric chemical transport models: the UK Met Office Numerical Atmospheric-235 dispersion Modelling Environment (NAME; Jones et al., 2007) Cantrell, 2008) to estimate the emissions of the trace gases by ISC method. A recent study  suggested that Weighted Deming Regression (WDR; hereafter, DR) method estimates a relatively more accurate slope and intercept by minimizing the residual errors for both X and Y among the various linear regression methods, particularly for atmospheric data with measurement error. As mentioned earlier, the 250 calculated slopes can be different depending on which linear regression fit is used. Therefore, we applied not only the DR approach but also the Fitexy and WYR methods to determine annual slopes between the observed enhancements of CH3Br and CFC-11 during 2008 to 2019. The results for the Fitexy and WYR methods are similar. Even though the co-matched observation points were slightly scattered in the range of large enhancements, the DR generated best fits representing the overall correlations trends. Millet et al. (2009) required a Pearson correlation coefficient (R) over 0.3. In order to distinguish 255 the contamination due to natural sources of CH3Br, and consider the origin of anthropogenic sources, we used only the data in which CH3Br and CFC-11 enhancement occurred at the same time for linear regression. Figure 7 shows the resulting annual slopes. For most of the observations, CH3Br enhancements show a correlation with those for CFC-11 with R larger than 0.4 (e.g., typically, R = 0.48 in 2011). They do not maintain a high correlation (R > 0.4) for every single year since most of the enhancements of CH3Br and CFC-11 were less than 5 ppt and high pollution events occurred only occasionally within a year. 260 Note that R in 2019 was very low (R <0.1) because of a tendency for the data in 2019 to bifurcate due to the occurrence of some high concentration cases from different source regions to the source regions for the majority of the low enhancement concentrations. For 2019, we adopted the slope and uncertainty of the regression line in 2010, which were used to estimate the emissions of CH3Br for 2019 by using the ISC method. Nevertheless, CFC-11 seems suitable as a reference compound to trace anthropogenic emissions from eastern China. Further, in general, if outliers are included in the analysis within the regression 265 process, R may not be robust and the regression slope may be heavily biased by the outliers (Devlin et al., 1975). Therefore, we applied robust WDR that can cover the overall scatter trend well, and it demonstrated that there was no significant difference between the regression results using all observation data and the outliers removed (See Fig. S4). In addition, the WDR slopes are well consistent with the annual medians of the individual ratios between ∆CH3Br and ∆CFC-11 data (See Table S1), which are known to be less sensitive to outliers compared to the means (Miller et al.,2012), implying that the resulting slopes are 270 robust to outlier data points and represent well the individual ratios between CH3Br and CFC-11 enhancements, as well.  Table S2; Carpenter and Reimann et al., 2014;TEAP, 2020).  In the life cycle of rapeseed, CH 3 Br is largely emitted during the flowering period in the 2 months after sowing (Jiao et al., 2020). Rapeseed in the northern hemisphere generally blooms in the warm weather from March to May. So seasonal emissions from the arable land of rapeseed may be related to the observed springtime increase in CH3Br polluted mole fractions at Gosan (See Figure S5). 315

Estimated CH3Br emissions from eastern China
China is the third-largest producer of rapeseed in the world after the European Union and Canada, accounting for 12% of the total rapeseed production in 2015-2016, and the arable land lies mainly along the Yangtze River, which is suitable for growing rapeseed (Khir et al., 2017). Previous studies have reported that the global emissions of CH3Br by the rapeseed industry range from 2.8 ± 0.7 Gg yr -1 (Jiao et al., 2020) to 5 Gg yr -1 (Gan et al., 1998;Mead et al., 2008). Considering the proportion of eastern China in the global rapeseed industry, the emissions of CH3Br by rapeseed in eastern China could be about 0.3-0.6 Gg yr -1 . 320

(ii) Biomass burning of agricultural residues
Owing to the almost total phase-out of CH3Br for non-QPS uses to date, the largest contributor to global anthropogenic emissions of CH3Br is biomass burning, such as agricultural open-field burning and use of biofuels (about 23 Gg yr -1 ; Carpenter and Reimann et al., 2014). As shown in Figure 6, the elevated mole fractions of VOCs (toluene, benzene, ethane), which are associated with biomass burning, are correlated with elevated mole fractions of CH3Br, suggesting that there may be some 325 contribution of biomass burning to the observed CH3Br enhancements. Note, the sources of VOCs pollution are generally not entirely due to biomass burning, as VOCs are emitted by combustion processes in general (e.g., fossil fuel use and combustion).
Approximately 140 Tg of agricultural residues are burned in fields across all of China every year (Zhao et al., 2017). Biomass burning in eastern China is predominantly due to the burning of agricultural crop residues (~60 Tg yr -1 ), mainly wheat residues (in May-June), rice and corn residues (in September-October) (Zhang et al., 2020). This eastern China biomass burning 330 seasonality may contribute slight partly to the seasonality in elevated levels of CH3Br seen at Gosan (May-June and September-October, see Figure S5).
The global annual emissions of CH3Br from the burning of agricultural waste are uncertain. Recently, Andreae (2019) has revised the emission factor (EF) of CH3Br by agricultural residues based on a field experiment to 1.1 g tonnes -1 of dry matter burned, and based on this, the global biomass burning emission of CH3Br by agricultural residues estimates was 0.3 Gg yr -1 . 335 Using this EF, the emissions of CH3Br from biomass burning of agricultural residues in eastern China would be approximately 0.07 Gg yr -1 .

(iii) Post-harvest treatment
Historically, CH3Br consumption resulted from soil fumigation (non-QPS), structural fumigation (non-QPS) and post-harvest fumigation (mainly QPS). Currently, the phase-out of CH3Br has been successfully implemented under the Montreal Protocol 340 for non-QPS applications, in particular the decrease in consumption of CH3Br for soil fumigation. Chemicals (e.g., chloropicrin, metam sodium, dazomet, etc.) and non-chemical methods (steam, soilless culture, resistant varieties) have been successfully introduced as alternatives to CH3Br use as soil fumigants (Mao et al., 2016;MBTOC, 2018). For QPS applications, phosphine has been widely used as a substitute for CH3Br in post-treatment of commodities, but it is known that some pests have developed resistance to phosphine (Jagadeesan and Nayak, 2017;Xinyi et al., 2017). SO2F2 is used in China as an alternative 345 to non-QPS use of CH3Br for the pre-plant soil fumigation as well as the QPS disinfestation of some durable products and post-harvest commodities (Cao et al., 2014;Gressent et al., 2021). Interestingly, the spatial distribution of the potential emission source regions estimated from the SO2F2 pollution observed at Gosan is very similar to that for CH3Br ( Figure S6).
In addition, the mole fractions of SO2F2 and CH3Br increase contemporaneously, and the correlations between the enhancements of both substances and CFC-11 are significant ( Figure S7 and S8). This implies temporal and spatial co-350 emissions of SO2F2 with anthropogenic CH3Br into the atmosphere. Gressent et al. (2021) showed that SO2F2 emissions in China were predominantly generated by post-harvest treatment rather than structural fumigation among its main uses, and were distributed within a large portion in eastern China. It seems that CH3Br and SO2F2 use source was spatially colocated, thus they are not completely replaced and co-emitted with its jumbled usage. Using these Gressent et al. SO2F2 emissions,the CH3Br emissions from eastern China for post-harvest treatment derived by the ISC method from the observations of SO2F2 12 and CH3Br at Gosan were 0.9 ± 0.2 Gg yr -1 for the period 2014-2019 (Table S3). Thus, the post-harvest use of CH3Br in eastern China results in approximately 1 Gg yr -1 of anthropogenic CH3Br emissions.

(iv) Unreported or inaccurately reported emissions from fumigation usage
The CH3Br emissions proposed above in (i)-(iii) can account for about half of the discrepancy (2.9 Gg yr -1 ) between 'top down' and 'bottom up' estimates for east China. The sources of the remaining discrepancies (~1.4 Gg yr -1 ) in CH3Br emissions remain 360 unknown.
Errors in the reported inventory for regulated uses cannot be ruled out because it is unsure whether the limits on new QPS use have been adhered to (MBTOC, 2018). Besides, despite the successful reduction of anthropogenic CH3Br emissions globally, the possibility of unidentified sources of emissions has been raised in multi-year MBTOC assessment reports (Porter and Fraser, 2020). As a similar example, we note that, although CFC-11 was a very important target chemical for phase-out under 365 the Montreal Protocol, unexpected CFC-11 emission increases were found due to unreported production and use in eastern China during 2013-2018 (Rigby et al., 2019;. In addition, it may be premature to conclude that CH3Br non-QPS use in China has been completely replaced by the alternatives discussed above. Since CH3Br represents the lowest costeffective fumigation method, the transition to the use of alternatives may be delayed without strong regulations and/or financial incentives and/or social awareness. The fact that CH3Br emissions derived from atmospheric observations in this study are 370 significantly larger than reported emissions suggests that unreported fumigation use of CH3Br may have occurred during the transition to alternative fumigation methods or that other sources, such as emissions from industrial wastes, have been overlooked.

Summary and conclusion
Atmospheric CH3Br has both natural and anthropogenic sources and plays a significant role in stratospheric ozone destruction. 375 For this reason, CH3Br non-QPS uses as a soil, commodity treatment and structural fumigant are being phased-out globally under the Montreal Protocol on Substances that Deplete the Ozone Layer, and its QPS use as a commodity fumigant is regulated.
To understand the temporal trend in atmospheric CH 3 Br abundances and its emission sources in East Asia, we analyzed the mole fractions of CH3Br observed at Gosan (Jeju Island, South Korea) for 12 years from 2008 to 2019. The baseline mole fractions indicating the regional state of the background atmosphere have decreased by -0.13 ± 0.02 ppt yr -1 (-1.5 % yr -1 ) during 380 the period, with seasonal variations increasing in spring and decreasing in summer. Despite the decreasing trend of the CH3Br baseline, relatively constant-strength pollution events occurred in every year.
A statistical backward trajectory analysis showed that emissions of CH3Br in the region were highest from eastern China compared to other surrounding countries. Top-down emissions estimates of CH3Br from eastern China were determined by using an ISC method with CFC-11 as the reference tracer defining anthropogenic CH3Br emissions. The ISC-based CH3Br 385 emission rates were 4.1 ± 1.3 Gg yr -1 on average during 2008-2019 and, despite the CH3Br phase-out for non-QPS applications in Article 5 countries, which includes China, in 2015, significant CH3Br emissions have continued. These CH3Br emissions determined from atmospheric observations are significantly different from the bottom-up emission estimates predicted from consumption data reported to UNEP (1.1 ± 0.2 Gg yr -1 ). The possible contributions of rapeseed industry and biomass burning to this discrepancy were assessed at approximately 0.3-0.6 Gg yr -1 and 0.07 Gg yr -1 , respectively. However, it is insufficient 390 to explain the approximate 3 Gg yr -1 difference between top-down (4.1 Gg yr -1 ) and bottom-up (1.1 Gg yr -1 ) estimates.
The remaining discrepancy (3.5 Gg yr -1 ) that ruled out the non-fumigation sources (rapeseed industry and biomass burning of agricultural residues) from total top-down CH3Br emissions is most likely due to fumigation use that was not reported and/or inaccurately reported or emissions from unknown sources, such as industrial waste or other sources. Correlations between CH3Br and SO2F2 pollution levels at Gosan suggest that the post-harvest use of CH3Br in eastern China contributes 0.9±0.2 395 Gg yr -1 to this 3.5 Gg yr -1 discrepancy. These data may suggest that the transition from CH3Br to SO2F2 or other alternatives for post-harvest fumigation in eastern China is only partially complete. Unreported use for fumigation may be related to the delay in introducing alternative technologies to CH3Br fumigation in east China and/or the lack of social awareness of the regulation, during the transitional period to alternative technologies.
Most of our estimated emissions of CH3Br are from eastern China and these CH3Br emissions, likely from unreported or 400 inaccurately reported fumigation usage, are significant enough to account for 30-40% of global emissions for fumigation usage. The total tropospheric bromine (in units of ppt) from long-lived brominated substances (CH3Br and halons) controlled by the MP has been decreasing since reaching a peak in 1998, mainly due to the decline of CH3Br. However, the contributions of halons to declining tropospheric bromine have become predominant since 2012 (Carpenter and Reimann et al., 2014). In recent years, CH3Br has been accounting for a significant proportion of the total amount of bromine in the troposphere from longlived compounds. Consequently, if any potentially unreported non-QPS and QPS emissions from fumigation usage could be 410 reduced and eventually stopped in developing countries, a further reduction of atmospheric CH3Br mole fractions would occur very quickly, due to the short half-life of CH3Br. For this reason, continued monitoring of atmospheric CH3Br mole fractions in East Asia and improvements in inverse modelling approaches are presently seen as a key priority in order to locate and identify specific emission sources.

Author contributions
HC, SP, and PJF designed the study; HC, SP, PJF, IP, JM, and JK interpreted the analyzed results and wrote the manuscript;

Top-down and bottom-up estimates of anthropogenic methyl bromide emissions from eastern China
Haklim Choi 1 , Mi-Kyung Park 1 , Paul J. Fraser 2 , Hyeri Park 3 , Sohyeon Geum 3 , Jens Mühle 4 , Jooil Kim 4 , The regional distribution of potential CH3Br sources in East Asia was derived by applying a statistical analysis to the air mass trajectories that correspond to the enhancements of CH3Br above baseline at Gosan during 2008-2019. Among the various trajectory-based statistical approaches, we applied a trajectory statistics method based on Seibert et al. (1994) to identify the potential sources of atmospheric pollutants . This method assumes that the concentration enhancement above baseline at the observation site is proportional to the average concentration of each grid cell through which 25 the air mass has passed and the residence time that the air mass spends in each grid cell. Therefore, the residence-time-weighted mean concentration ̅̅̅̅̅ of a target compound in each grid on the domain can be calculated as follows: 30 where m,n is a potential source region of CH3Br -m,n are indices of a horizontal grid cell, i is the index of the trajectory and M is the total number of trajectories, is the enhanced concentration of CH 3 Br above baseline and is the residence time that trajectory i spent over the grid cell m,n within the atmospheric boundary layer. The calculation of the residence time over each grid was accomplished using the method of Poirot and Wishinski (1986), which assumes that an air parcel travels linearly between two points at constant speed. 35    . S2a shows the 17-potential source regions for East Asia. The regional origin of an air-mass that inflows to Gosan is 50 classified from the backward trajectory analysis using HYSPLIT. In this study, the aggregated source regions were designated as China (2, 3, 15, 16 and 17), eastern China (15, 16 and 17), Korea (8, 9 and 10) and the remining regions were classified as others (as shown in Fig. S2b). Eastern China-1 consists of Beijing, Liaoning, Tianjin, Hebei and Shanxi provinces, while eastern China-2 consists of Henan, Hubei, Anhui, Jiangsu, Shanghai, Jiangxi, Zhejiang and Fujian provinces. manuscript; the regional origin of each air mass is indicated by colour with regards to the 17 regions (Fig S2a), aggregated to eastern China, China, Korea and others.   Fig. 7 of main text; black) and removed outlier data (yellow). The yellow asterisks correspond to observations that were considered outliers and removed. Figure S4 shows the xy plot of the annual enhancements of CH3Br and CFC-11 above baseline for all data (black) and data with outliers removed (yellow). To filter out outliers, we selected data in the range of Q1-1.5*IQR < the difference of CH3Br and CFC-11 < Q3+1.5*IQR (outliers removed). The linear regression between the two pollutants were derived from the weighted Deming regression (WDR) method suggested . As described in Table S1, the estimated annual slopes and uncertainties between CH3Br and CFC-11 by WDR does not differ significantly with or without outliers. This 75 demonstrated the robust WDR that can cover the overall scatter trend well. 80 Table S1: Annual slopes and their uncertainties between ΔCH3Br and ΔCFC-11 as shown in Figure S4. The annual median values for individual ratios (ΔCH3Br/ΔCFC-11) along with 16 th and 84 th percentile ranges.    110 Figure S8: Same as Figure S4, but for CH3Br and SO2F2.  Table S1, but for CH3Br and SO2F2 as shown in Figure S8, and annual SO2F2 emissions and estimated CH3Br 115 emissions for post-harvest treatment by ISC method. The post-harvest treatment SO2F2 emissions were derived for eastern China from the global emission data of Gressent et al., 2021.