Changing characteristics of atmospheric CH 4 on the Tibetan Plateau , 1 records from 1994 to 2017 at Mount Waliguan station 2 3

Chinese Academy of Sciences, Beijing, China 7 2College of Resources and Environment, University of Chinese Academy of Sciences, Beijing, China 8 3College of Environment, Zhejiang University of Technology, Hangzhou, China 9 4Mt. Waliguan background station, China Meteorological Administration (CMA), Qinghai, China 10 5Meteorological Observation Center (MOC), China Meteorological Administration (CMA), Beijing, China 11 6College of Global Change and Earth System Science, Beijing Normal University, Beijing, China 12 7Institute of Ecology, Key Laboratory of Agrometeorology of Jiangsu Province, School of Applied 13

through the meteorological approach, which was based on essential meteorological 216 information, similar to previous studies by Zhou et al. (2004) and Liu et al. (2019). In 217 this study, the CH4 records associated with surface wind from selected sectors (i.e. 218 NNE-…-ENE in spring, NE-…-SE in summer, NE-…-ESE in autumn, and NE-ENE in 219 winter) were flagged as local representative. Subsequently, we further rejected portion 220 of daytime records to minimize the effect of human activities (e.g., rush hours), 221 including 9:00-13:00 LT (local time) in spring and summer, 9:00-14:00 LT in autumn, 222 10:00-17:00 LT in winter. Finally, we filtered CH4 data into local events when the 223 surface wind speed was less than 1.5 m s -1 to minimize the very local accumulation. 224 To understand the influence of local surface wind, the hourly CH4 data was 225 calculated versus 16 horizontal wind directions (Fang et al., 2013). In this study, we 226 used the 'polarPlot' function located in the 'openair' package of the statistical software 227 polar plot of CH4 concentrations, and the concentrations are calculated as a continuous 229 surface by modelling using smoothing techniques (Carslaw et al., 2006;Diederich, 230 exactly receptor site (Ashbaugh et al., 1985). In this study, PSCF value was calculated 244 in 0.5×0.5-degree grid cell (i, j): 245 PSCF / ij ij ij m n = (1) 246 nij represents the number of endpoints that terminate in the ijth grid cell, while the 247 number of trajectories with concentration exceed the threshold value was defined as mij 248 (Polissar et al., 1999). In order to reduce the abnormal influence of small nij values in 249 some grid cells, PSCFij was further computed by an arbitrary weighting function Wij as 250 Wij represents the weight of cell (i, j), nij is the number of trajectory endpoints that fall 253 in the ijth grid cell, while the nave shows the mean number of the endpoints in all grid 254

cells. 255
In order to fill the data gaps so as to evaluate the long-term CH4 trend, we applied 256 the curve fitting approach by Thoning et al. (1989). We also calculated the trend curve 257 that excluded the influence of seasonal variation, and then got the annual growth rates 258 of the average of the first derivative of the trend curve. The function consists of the 259 polynomial part and the annual harmonics part: 260 CH4 reached at 9:00-13:00 LT with one peak at noon. In winter, largely increasing 296 presented in the daytime at 9:00-17:00 LT, with the largest amplitude of 6.2 ± 2.4 ppb 297 among four seasons. For the diurnal variation over the whole monitoring period, the 298 highest CH4 mole fraction was observed in winter and the minimum value was found 299 in spring (Fig. 2f). 300 Different patterns for diurnal CH4 cycles were also found over different periods. 301 In 1994-1997 and 1998-2002, the CH4 mole fractions in winter were apparently higher 302 than the other seasons ( Fig. 2a-b). But its value was the highest in summer during the 303 period of 2003-2007, 2008-2012and 2013. The atmospheric CH4 304 values in winter were gradually falling behind the other seasons, and the gaps among 305 different seasons were increasing, especially for summer. Before 2002, diurnal cycles 306 in four seasons were ambiguous ( Fig. 2a-b), but significant diurnal variations appeared 307 afterwards ( Fig. 2c-e). The peak to trough amplitude almost increased along with the 308 time in almost all seasons. For example in spring, the amplitude was 6.5 ± 3.0, 4.7 ± 309 1.8, 5.6 ± 2.6, 6.2 ± 2.4 and 6.8 ± 3.4 ppb over 1994-1997, 1998-2002, 2003-2007, 310 2008-2012 and 2013-2017, respectively (Table S1). 311 312 3.2 The impact of local surface winds 313 As observed by the previous short-term variations, the atmospheric CH4 at WLG was 314 significantly influenced by local surface wind from northeast to southeast sectors (Fig.  315 3f). Slight differences were also found among seasons. In spring, the atmospheric CH4 316 was enhanced by 2.5-6.5 ppb compared to the seasonal average (1839.7 ± 1.4 ppb) 317 when the wind was originating from NNE-NE-ENE-E sectors. In summer and autumn, 318 the wind from NNE-NE-ENE-E-ESE induced higher CH4 mole fractions, with 319 enhancement of about 3-9.5 ppb and 4 to 18 ppb, respectively. In winter, the CH4 mole 320 fractions significantly elevated from the wind sectors that same as those found in spring, 321 with value of 7-21 ppb than seasonal average (1854.5 ± 4.8 ppb). Relatively, the 322 amplitude of enhancements in winter and autumn were larger than those in spring and summer. 324 What interesting is that wind sectors elevating CH4 mole fractions vary in different 325 periods. The early periods (i.e. 1994-1997 and 1998-2002) were different from the 326 recent periods (2003-2007, 2008-2012 and 2013-2017). The elevated CH4 was 327 predominately from about ENE-E-ESE-SE-SSE sectors in early years ( Fig. 3a-b), but 328 evolved to NNE-NE-ENE-E sectors in later years . Furthermore, the 329 amplitude of enhancements was almost increasing continuously along with the time. 330 For example, in autumn, the maximum CH4 mole fractions were from E, ENE, ENE, 331 NE and ENE sectors in 1994-1997, 1998-2002, 2003-2007, 2008-2012and 2013 with the successively increasing enhancements of 8. 6, 12.1, 14.7, 16.8 and 19.7 ppb, 333 respectively. 334 We applied the CPF to hourly CH4 and CO data by considering intervals of entire 335 data percentiles including 0-20, 20-40, 40-60, 60-80 and 80-100 to draw the CPF polar 336 plot. It is clear that different sources only affected CH4 mole fractions on different 337 percentile range. For example, for most wind speed-directions the CPF probability of 338 CH4 being greater than the 60 th percentile was tending to zero (Fig. 4). And it is apparent 339 that most sources contributed to the less than 60 th percentiles of CH4 mole fraction (e.g. 340 40-60) (Fig. 4). The specific sources were prominent for specific percentile ranges. The 341 wind from the southwest and southeast was important on the cases of the higher 342 percentiles, resulted in the highest CH4 mole fractions of 1849-1872 ppb for 60-80 343 percentile and 1872-2031 ppb for 80-100 percentile (Fig. 4). It's more obvious that the 344 CO showed gradually shifted sources with the increase of percentile ranges. The areas 345 where the CPF probabilities were higher is to the NW-SW sectors when the percentages 346 ranged from 0 to 40 th . Nevertheless, when the percentages were larger than 60 th , the 347 high probability areas completely moved to NE-SE sectors (Fig. 4). regions, which accounted for about 24% (cluster 3) and 44% (cluster 5) of total 354 trajectories (Fig. 5a). These air masses were also accompanied with higher CH4 mole 355 fractions than those from east and northeast regions, i.e. cluster 1 (13.3% of total) and 356 cluster 4 (11.69%) ( Table 1). The largest enhancement was ~18 ppb (relative to spring 357 average) by cluster 3. In summer, 45% of air masses (cluster 1) were from eastern 358 regions. But the high CH4 mole fractions were on cluster 2 and cluster 5 from northwest 359 and west regions, although low percentages were found (cluster 2: 26%; cluster 5: 7%) 360 ( Fig. 5b). The highest CH4 mole fraction was associated with cluster 2, with ~9 ppb 361 larger than the average in summer. In autumn, large proportion of air masses was 362 originating from west and southwest station, such as cluster 2 (49%) and cluster 3 (32%) 363 ( Fig. 5c). The highest CH4 was from cluster 3 with enhancement of ~4 ppb than the 364 seasonal average. Similar to autumn, the air masses were primarily from northwest in 365 winter, including northwest cluster 3 (59%) and southwest cluster 1 (34%) regions ( Fig.  366 5d). The highest CH4 mole fractions was on cluster 1with the enhancement of ~7 ppb 367 than the average value. 368 369

Spatial distribution of potential source regions 370
In this study, the potential sources were analyzed over different periods, i.e. 2004-2007, 371 2008-2012 and 2013-2017. Generally, the strong potential sources were located at the 372 northeast to southeast of the station, especially in summer, but the source regions 373 differed in various seasons as well as years (

Extracting the well-mixed ambient methane 384
To precisely understand characteristics of atmospheric CH4, e.g., seasonal cycle or 385 long-term trend, it is vital to identify the CH4 records that were influenced by local 386 sources and sinks. In this study, we analyzed hourly CH4 measurements during 1994-387 2017, and 47.3% of CH4 data were classified as regional representative, with the 388 average CH4 mole fraction of 1847.9 ± 0.3 ppb. The local representative data was 389 obviously larger than regional events, with an average value of 1858.2 ± 0.4 ppb (Table  390 2). The proportion of regional events increased slightly before 2012, but significantly 391 reduced in recent years (e.g., 2013-2017). The filtered regional/local time series was 392 shown in Figure 7. It can be seen that the CH4 mole fractions obviously increased from 393 1994 to 2017. The atmospheric CH4 showed strong growth and displayed large 394 fluctuation at WLG (Fig. 7). In 1995, the average CH4 mole fraction was only 1805.8 395 ± 0.1 ppb, however, the average value increased 98 ppb by the year of 2016 (1903.8 ± 396 0.1 ppb) (Table 3). 397 In order to further investigate the characteristics of atmospheric CH4, we divided the 414 CH4 observations into two main regions according to the above analysis, including 415 geographical conditions, the effect of surface winds, the long-range transports and the 416 potential source distributions. The first region was covered the northeast to southeast 417 (NNE-…SE) of WLG, which was denoted as City Regions (CR). The second region 418 was located south to west (S-…-W) of the station and was well known Tibet (Qinghai-419 Xizang) Plateau (TP) (Fig. S2). Accordingly, the hourly CH4 records when the surface 420 wind coming from these sectors were divided into two subsets (i.e. TP and CR). The 421 long-term variations between the two regions as well as the total regional time series 422 (Total) were compared and analyzed to explore new sight of atmospheric CH4 variation 423 at WLG. 424 Overall, at WLG, the seasonal averages of CH4 were ordered by summer (1850.0 425 ± 0.3 ppb), winter (1847.4 ± 0.3 ppb), autumn (1844.4 ± 0.3 ppb) and spring (1841.2 ± 426 0.3 ppb), except during 1994-1997 with the maximum in winter and minimum in 427 autumn (Fig. 10). Seasonal averages in CR were significantly different to that in TP and 428 also the entire regional data (Total). The seasonal average in TP was mostly higher than that in CR, except for wintertime (Table S2). The atmospheric CH4 in August was 430 mostly the maximum and the April was the minimum for the total regional time series 431 (Total), with the seasonal amplitude of 13.4 ppb. The peak to trough amplitude in CR 432 (~15 ppb) was higher than that in TP (~13 ppb) during 1994-2016. Additionally, 433 seasonal amplitudes indicated different trends between CR and TP. For CR, the seasonal 434 amplitude was firstly dropped and then increased along with time, which were similar 435 to the variation of total regional events (Total). But for TP, the amplitude displayed a 436 continuously increasing trend, with values of about 15.9, 19.3, 21.6, 23.4 and 22.4 ppb 437 in 1994-1997, 1998-2002, 2003-2007, 2008-2012 and 2013-2016, respectively. 438 439 3.6.2 Long-term trend 440 In the 1990s, the CH4 growth rates were very low and even negative at WLG. 441 Subsequently, during 2002-2006, a steady period was found with a near-zero growth 442 rates. After 2007, the atmospheric CH4 was raised significantly (Fig. 11a). In the year 443 of 1997/1998, 2000/2001, 2007/2008 and 2011/2012, larger amplitude of the growth 444 rates was found and strong growth appeared (Fig. 11b) (red color) (Fig. 11). 451 The overall annual growth rates were 5.1 ± 0.1 ppb yr -1 over 1994-2016 at WLG 452 (Table 4). However, the periodic annual growth rats were 4.6 ± 0.1, 2.6 ± 0.2, 5.3 ± 0.2, 453 7.6 ± 0.2 and 5.7 ± 0.1 ppb yr -1 in 1994-1997, 1998-2002, 2003-2007, 2008-2012 and 454 2013-2016, respectively. The CH4 growth rate in CR was significantly different from 455 that in TP (Fig. S3). In 1994In -1997In , 2003In -2007In and 2013, the growth rates in TP 456 were obviously larger than that in CR (Table 4). But in 2003-2007 and 2008-2012, the previous studies on black carbon (BC) and carbon monoxide (CO) indicated that the 509 emissions from the Yellow River Canyon industrial area, ~500km away from northeast 510 of WLG may also donate to the high CH4 values originating from ENE and NE sectors 511 (Zhou et al., 2003). In summer, the prevailing wind directions were from NE-…-ESE 512 sectors (~46%) (Fig. S6), and the CH4 mole fractions were also higher in the related 513 sectors. However, in the autumn and winter, although the prevailing wind and high wind 514 speed were from SSW-…-W sectors (~ 40-50%) (Fig. S6), the high CH4 mole fractions 515 were from almost the opposite wind sectors of NNE-….-ESE (Fig. 3f), which indicated 516 that strong local sources were distributed from northeast to southeast (city regions), and 517 even covered the emissions of natural sources. As time goes on, the wind sectors with 518 high CH4 mole fractions changed and concentrated on ESE to ENE sectors, and the 519 amplitudes of enhancements were increasing, which further implied the effect of 520 stronger emissions from anthropogenic sources in city regions in recent years. 521 The air masses from east and northeast regions passed over the cities of Xining and 525 Lanzhou (capital of Gansu province), which is the populated center and industrial area 526 (Fig. 5). However, the highest CH4 values was not observed when air mass was from 527 these sectors. Instead, high CH4 mole fractions were frequently observed when air mass 528 from the northwest to southwest (Table 1). It was possibly due to that the air masses 529 from west and northwest had passed through the northwest of Qinghai province and the 530 central area of Xinjiang Uygur Autonomous Region (XUAR), where located Ge'ermu 531 urban area (the second largest city of Qinghai) with rapid industrial development, 532 natural gas and petroleum resources exploitation and large crops residue burning 533 (Zhang et al., 2013). Similar to the CPF percentile analysis (Fig. 4), the southwest or 534 northwest region away from the site may be also strong source regions. 535 https://doi.org/10.5194/acp-2020-481 Preprint. Discussion started: 29 July 2020 c Author(s) 2020. CC BY 4.0 License.
Most potential source identified in northwestern regions (Fig. 6) was possibly due 536 to CH4 emissions from the northwest Gansu province, the northwest Qinghai province 537 and the southeast of XUAR. The different source distribution by seasons could be 538 attributed to the effect of westerlies or the southeast monsoons (Zhou et al., 2004). The 539 obviously increasing source region was clear evidence for the strong effect of the 540 expansion of human activity. Moreover, the pattern of source region moved from the 541 east to the southwest, especially in autumn and winter, indicated that the southwest 542 away from the WLG, e.g., Northern India, were gradually becoming a strong CH4 543 source region. India has abundant cattle as well as extensive large-scale coal mining, 544 large amount of CH4 emissions may transport from northern India to the northeastern 545 Tibetan Plateau (Fig. 6i & l). The air mass transport result (Fig. 5d)

Different sources between CH4 and CO 551
The percentile polar plot clearly showed the specific distribution of different CH4 mole 552 fractions (Fig. 4). The result revealed that most areas around WLG contributed to low 553 CH4 mole fractions, the southeast and southwest of the site exist two strong source 554 regions. It is of great possible that the anthropogenic emission from cities (e.g., 555 Lanzhou, Chengdu, etc.) was the only cause for high values in the southeast, and the 556 southwest region away from WLG was possibly due to sources from other countries, 557 such as India. Unlikely to the CH4, the high CO mole fractions were consistent from 558 east regions (urbanized area) (Fig. 4), indicating strong anthropogenic sources in city 559 regions (i.e. Xining and Lanzhou) (Zhang et al., 2011). 560 The seasonal cycles of ΔCO/ΔCH4 slopes (high in summer and low in winter) (Fig.  561 8) may primarily be due to the effect of monsoons and air mass transport. In summer, 562 air masses arriving at WLG were predominantly transported from the northeast to east city regions (e.g., Xining, Lanzhou) with the largest CO mole fractions. In contrast, the 564 air masses were mainly from the southwest in winter, which carried strong CH4 565 emissions but few CO emissions (Zhang et al., 2011) (Table 1). Hence the opposite two 566 air mass transport lead to a peak in summer and a trough in winter (Fig. 8). Moreover, 567 we could see apparently regional polarization in the concentration ratio of CH4 and CO 568 ( Fig. S7), implying the different strong source distribution between CH4 and CO at 569 WLG. The cluster results (Fig. 5b & d) and the potential sources analysis (Fig. 6f & l) 570 also support this seasonal variation. Tohjima et al. (2014) found an opposite variation 571 at Hateruma Island, which showed low slope values in summer. It could be attributed 572 to different local sources and sinks, suggested the special topography condition and 573 local source distribution around WLG. The large ΔCO/ΔCH4 fluctuations (Fig. 9) over 574 the study period was likely because of the anomaly years of different CH4 or CO mole 575 fractions as well as source regions. In 2007, large increase of ΔCH4 appeared, and from 576 2010 to 2013, the ΔCO decreased significantly (Fig. S1). Before 2010, large air masses 577 and potential source regions were identified in eastern regions (city regions) with the 578 highest CO emission ( Fig. 5; Fig. 6). After 2010, the southwest regions showed high 579 contributions, with the highest CH4 emission but relatively lower CO emission. 580 Therefore, obviously variation of the slopes presented with almost an increase in 2007-581 2010 and a decrease after 2010 (Fig. 9). 582 583

Long-term trends in different observing periods 615
The entirely fluctuant trend of atmospheric CH4 in 1994-2016 at WLG (Fig. 11) was 616 similar to the global trend reported by quite a few studies (Bergamaschi et al., 2013; the CH4 annual increase of 4.5 ppb yr -1 in 1992-2001, which was similar to that in 1994-619 1997 (4.6 ± 0.1 ppb yr -1 ) and 1997-2002 (2.6 ± 0.2 ppb yr -1 ) in our study (Table 4). 620 Tohjima et al. (2002) found similar growth rates that the CH4 at Cape Ochi-ishi and 621 Hateruma Island in1995-2000 was increased about 4.5 and 4.7 ppb yr -1 , respectively. In 622 early 1990s, the CH4 trend at WLG is very low, which was similar to the global growth 623 rates (Fig. 11b). The levels of OH radicals may have controlled the decrease or increase 624 of CH4 in the atmosphere during this period (Dlugokencky et al., 1998;Rigby et al., 625 2017;Turner et al., 2017). The growth rates were high in 1998 (Fig. 11b), which may 626 have been due to the high temperatures, large biomass burning and weak destruction 627 (Cunnold et al., 2002;Lelieveld et al., 2004;Simmonds et al., 2005). 628 The continuously larger CH4 growth rates after 2007 at WLG (Fig. 11), e.g., 7.6 ± 629 al., 2016). The study by Chen et al. (2013) illustrated that the warming (0.2 °C per 675 decade) in the Tibetan Plateau resulted in substantial emission of CH4 due to the 676 permafrost thawing and glaciers melting. However, up to now, the specific causes of 677 such distinct variability around the years, e.g., the spikes or near-zero CH4 growth rates, 678 are not yet determined. 679 680

Annual growth rate in Qinghai-Tibetan Plateau 681
Although similar annual growth rates were found among the City Regions (CR), the 682 Tibet Plateau (TP) and total regional records (Total) in the entire observing period 683   (Fig. 11), significant differences were found in short-term periods (Fig.  684   S3). In 2013-2016 (Table 4), the TP showed larger growth rate than that in CR, implying 685 Plateau, it may provide us one of the last precious regions to study global climate 703 changes (Chen et al., 2013). The anomalously year to year fluctuations of atmospheric 704 CH4 in Tibetan Plateau were unquestionably a warning or alarm to the world, and the 705 unprecedented annual growth rate might be a dangerous signal to global climate change. 706 707 5 Conclusion 708 The atmospheric CH4 at Mt.Waliguan increased continuously during 1994-2017. 709 Although near-zero and even negative growth appeared in some particular periods, e.g., 710 1999-2000, and 2004-2006, the overall trend of CH4 was increased rapidly, especially 711 in recent decade. Obvious diurnal cycle was found with the peak at noon and a trough 712 at late afternoon. Due to the unique geophysical locations and transport pathway, the 713 seasonal averages of CH4 at WLG displayed an opposite trend with sites in the northern 714 hemisphere, with summer maximum and spring minimum. Large amount of air masses 715 was from west and northwest regions of WLG, which accompanied with higher CH4 716 mole fractions than that from city regions. The Northern India possibly became a strong 717 source of CH4 to WLG rather than city regions before. 718 As time goes by, the temporal patterns (e.g., seasonal amplitude), the annual 719 variations, the long-term trends or potential source distribution of CH4 at WLG are all 720 changing. Thus, the long-term verification is extremely important to accurately 721 understand CH4 variations. The case study in Qinghai-Tibetan Plateau revealed 722 unprecedented annual growth rates of CH4. In recent years, the Tibetan Plateau even 723 showed larger growth rate than that in city regions. Tibetan Plateau was with the highest 724 average altitude and was almost impervious to strong human activities. There is no 725 doubt that the anomalously variation and the unprecedented annual growth rate of 726 of a large international airport, Atmos. Environ., 40, 5424-5434, 782 10.1016Environ., 40, 5424-5434, 782 10. /j.atmosenv.2006Environ., 40, 5424-5434, 782 10. .04.062, 2006 Chen  1974, J. Geophys. 984 Res.-Atmos., 94, 8549-8565, 10.1029/JD094iD06p08549, 1989 Tsutsumi, Y., Mori, K., Ikegami, M., Tashiro, T., and Tsuboi, K.: Long-term trends of 986 greenhouse gases in regional and background events observed during 1998-2004 at 987 Yonagunijima located to the east of the Asian continent, Atmos. Environ., 40, 5868-988 5879, 10.1016Environ., 40, 5868-988 5879, 10. /j.atmosenv.2006Environ., 40, 5868-988 5879, 10. .04.036, 2006 Uria-Tellaetxe, I., and Carslaw, D.     is the annual growth rates of atmospheric CH4 records from CR, TP as well as the total 1125 regional time series (Total). The growth rates are calculated from the first derivative of trend curves.

1126
The smooth curve and the trend is calculated by the method of Thoning et al. (1989).