Quantifying variability, source, and transport of CO in the 1 urban areas over the Himalayas and Tibetan Plateau

: 29 Atmospheric pollutants over the Himalayas and Tibetan Plateau (HTP) have potential 30 implications for accelerating the melting of glaciers, damaging air quality, water sources and 31 grasslands, and threatening climate on regional and global scales. Improved knowledge of the 32 variabilities, sources, drivers, and transport pathways of atmospheric pollutants over the HTP is 33 significant for regulatory and control purpose. In this study, we quantify the variability, source, and 34 transport of CO in the urban areas over the HTP by using in situ measurement, GEOS-Chem model 35 tagged CO simulation, and the analysis of meteorological fields. Diurnal, seasonal, and interannual 36 variabilities of CO over the HTP are investigated with ~ 6 years (January 2015 to July 2020) of 37 surface CO measurements in eight cities over the HTP. Annual mean of surface CO volume mixing 38 ratio (VMR) over the HTP varied over 318.3 ± 71.6 to 901.6 ± 472.2 ppbv, and a large seasonal 39 cycle was observed with high levels of CO in the late autumn to spring and low levels of CO in 40 summer to early autumn. The diurnal cycle is characterized by a bimodal pattern with two 41 maximums in later morning and midnight, respectively. Surface CO VMR from 2015 − 2020 in most cities over the HTP showed negative trends. The IASI satellite observations are for the first 1 time used to assess the performance of GEOS-Chem model for the specifics of the HTP. The GEOS- 2 Chem simulations tend to underestimate the IASI observations but can capture the measured 3 seasonal cycle of CO total column over the HTP. Distinct dependencies of CO on a short lifetime 4 species of NO 2 in almost all cities over the HTP were observed, implying local emissions to be 5 predominant. By turning off the emission inventories within the HTP in GEOS-Chem tagged CO 6 simulation, the relative contribution of long range transport was evaluated. The results disclosed 7 that transport ratios of primary anthropogenic source, primary biomass burning (BB) source, and 8 secondary oxidation source to the surface CO VMR over the HTP varied over 35 to 61%, 5 to 21%, 9 and 30 to 56%, respectively. The anthropogenic contribution is dominated by the South Asia and 10 East Asia (SEAS) region throughout the year (58% to 91%). The BB contribution is dominated by 11 the SEAS region in spring (25 to 80%) and the Africa (AF) region in July – February (30 – 70%). 12 This study concluded that main source of CO in urban areas over HTP is due to local and SEAS 13 anthropogenic and BB emissions, and oxidation sources, which differ from the black carbon that is 14 mainly attributed to BB source from Southeast Asia. The decreasing trends in surface CO VMR 15 since 2015 in most cities over the HTP are attributed to the reduction in local and transported CO 16 emissions in recent years.


Introduction 18
The Himalayas and Tibetan Plateau (HTP), also named the 'Third Pole' (TP), is an important 19 region for climate change studies due to several reasons. Due to its unique feature for interactions 20 among the atmosphere, biosphere, hydrosphere, and cryosphere, the HTP is referred to as an 21 important indicator of  35 Since the population level is very low, the HTP has long been regarded as atmospheric 36 background with negligible local anthropogenic emissions (Yao et al., 2012;Kang et al., 2019). 37 However, the HTP is surrounded by East Asia and South Asia which include many intensive 38 anthropogenic and natural emission source regions (Zhang et  2019b). 13 The ecosystem over the HTP is sensitive and fragile under the extreme alpine conditions. These 14 exogenous and local atmospheric pollutants have potential implications for accelerating the melting 15 of glaciers, damaging air quality, water sources and grasslands, and threatening climate on regional 16 and drivers, and transport pathways of atmospheric pollutants over the HTP are still not fully understood. 28 CO is one of the most critical atmospheric pollutant which not only threatens human health but 29 also plays a vital role in atmospheric chemistry (Zhang et  with hydroxyl radicals (OH). Since CO is heavily involved in the relationship between atmospheric 39 chemistry and climate forcing, it is crucial to investigate its atmospheric burden, variability, and 40 potential driver over the HTP. CO over the HTP may originate from various source regions and 41 sectors, improved knowledge of their relative contributions to CO variability over the HTP is also 42 significant for regulatory and control purpose. Furthermore, an investigation of CO pollution can 43 complement current atmospheric investigation over the HTP since the chemical characteristic, 44 climate forcing, and deletion of CO is different from the well-established carbonaceous aerosols. 1 In this study, we quantify the variability, source, and transport of CO in the urban areas over 2 the HTP by using in situ measurement, GEOS-Chem model tagged CO simulation, and atmospheric 3 circulation pattern techniques. Diurnal, seasonal, and interannual variability of CO over the HTP 4 are investigated with multiyear time series of surface CO measurements in eight cities over the HTP. 5 The performance of GEOS-Chem full-chemistry model for the specifics of the HTP is first assessed 6 with the concurrent satellite observations. The GEOS-Chem model is then run in a tagged CO mode 7 to quantify relative contribution of long range transport to the observed CO variability over the HTP. 8 The three-dimensional (3D) transport pathways of CO originated in various source regions and 9 sectors to the HTP are finally determined by the GEOS-Chem simulation, back trajectories analysis 10 and atmospheric circulation pattern. Only few studies have investigated the burden and variability 11 of CO over the HTP (Ran et al., 2014;Yin et al., 2019a). These studies uniformly focused on the 12 most developed regions in Lhasa, and did not analyze interannual trends and transport of CO. This 13 study not only expands the coverage of CO quantification over the HTP, but also provides insights 14 into the interannual trends, sources, and transport of CO in all urban areas over the HTP. 15 The next section describes site description, the surface in situ CO data and auxiliary data, the 16 methodology used to estimate the interannual trend of surface CO, and the GEOS-Chem simulation 17 used for source attribution. Section 3 reports the results for surface CO variability over the HTP on 18 different time scales. Section 4 presents the results for GEOS-Chem model evaluation. Section 5 19 analyzes the results for source attribution using GEOS-Chem tagged CO simulation and the analysis 20 of meteorological fields. We conclude the study in Section 6. 21 2 Methods and data 22 2.1 Site description 23 Surface in situ CO measurements in eight cities over the HTP are used in this study. The 24 locations of these cities are shown in Fig. 1 and summarized in Table 1. Ngari is located in the 25 western, Diqing and Qamdo is located in the eastern, and the rest cities are all in central eastern of 26 the HTP. Ngari, Shigatse, Lhasa, Shannan, and Nyinchi are adjacent to the Himalayas region, and 27 Naqu, Qamdo, and Diqing are relatively distant from the Himalayas region. Generally, these cities 28 represent the most developed and populated areas over the HTP. The altitude of these cities ranges 29 from 3.1 to 4.5 km a.s.l. and the population ranges from 110 to 770 thousand. The surface pressure 30 of these cities is about 600 hPa or less throughout the year (Table 1). Typically, all these cities are 31 formed at flat valleys with the surrounding mountains rising to more than 5.0 km a.s.l., and keep 32 continuous expansion and development over time. These cities are characterized by a typical climate 33 regime in high mountain regions, and is dry and cold in most of the year. Due to the high altitude 34 and thin air, incident solar radiations over these cities are stronger than those over other cities at the 35 same latitude around the globe (Ran et al., 2014). 36 General atmospheric circulation over these cities are typically influenced by three synoptic 37 systems: the warm and wet air masses during the monsoon season in summer, the South Asian 38 anticyclone that controls the upper troposphere and above, and the subtropical mid-latitude 39 westerlies in winter (Yao et The CNEMC network has monitored six surface air pollutants (including CO, O3, NO2, SO2,  5 PM10, and PM2.5) at 23 sites in eight cities in Ngari, Lhasa, Naqu, Diqing, Shigatse, Shannan, 6 Nyingchi, and Qamdo over the HTP (Table 1). Each city has at least two measurement sites. Surface 7 CO volume mixing ratio (VMR) measurements at all sites are based on similar gas correlation filter 8 infrared analyzers (http://www.cnemc.cn/en/, last access: 22 March 2020). The hourly mean datasets 9 have covered the period from January 2015 to present for all measurement sites in the eight cities 10 ( Table 1). We first applied filter criteria following that of (Lu et al., 2019) to remove unreliable 11 measurements. The resulting measurements at all measurement sites in each city are then averaged 12 to obtain a city representative dataset. 13 The 3D  profiles and cloud fractions used in FORLI are those from the EUMETSAT Level 2 processor. Only 32 pixels associated with cloud fraction below 25 % are processed. The IASI CO product is a vertical 33 profile given as partial columns in moles per square meter in 18 layers between the surface and 18 34 km, with an extra layer from 18 km to the top of the atmosphere. The pressure levels associated 35 with retrieval layers are provided with the CO product. This IASI CO dataset also includes other 36 relevant information such as a general quality flag, the a priori profile, the total error profile, the air 37 partial column profile, and the averaging kernel (AK) matrix, on the same vertical grid, and the total 38 column and the associated total error. To balance the accuracy and the number of valid data over 39 HTP, the IASI data within ±1° latitude/longitude rectangular area around each city and with total 40 error of less than 15% are selected. 41 42 We have used a bootstrap resampling model to determine the seasonality and interannual 43 respectively. A0 is the intercept, A1 is the annual growth rate, and A1/A0 is the interannual trend 10 discussed below. In this study, we incorporated the errors arising from the autocorrelation in the 11 residuals into the uncertainties in the trends following the procedure of (Santer et al., 2008). The A2 12 -A5 parameters describe the seasonal cycle, t is the measurement time elapsed since January 2015, 13 and ε(t) represents the residual between the measurements and the fitted results. Fractional 14 differences of measured CO VMR time series relative to their seasonal mean values represented by 15

Regression model for CO trend
( ) were referred to as seasonal enhancements and were calculated as equation (3). 16 17 Two types of GEOS-Chem model simulations were involved in this study. GEOS-Chem model 18 version 12.2.1 (DOI:10.5281/zenodo.2580198) was first ran in a standard full-chemistry mode to 19 be evaluated by the IASI CO product. The GEOS-Chem model was then ran in a standard tagged 20 CO mode to quantify relative contribution of long range transport to the observed CO variability January 2015) to remove the influence of the initial conditions. We only considered CO simulations 35 for the grid boxes containing the eight cities over the HTP. calculated by the resistance-in-series algorithm (Wesely, 1989;Zhang et al., 2001). The photolysis 6 rates were obtained from the FAST-JX v7.0 photolysis scheme (Bian and Prather, 2002). A universal 7

GEOS-Chem simulation
tropospheric-stratospheric Chemistry (UCX) mechanism was implemented (Eastham et al., 2014). 8 In GEOS-Chem tagged CO simulation, the improved secondary CO production scheme of 9 (Fisher et al., 2017) was implemented, which adopts secondary CO production rates from CH4 and 10 NMVOCs oxidation. The monthly mean OH fields and secondary CO production rates from CH4 11 and NMVOCs oxidation are archived from the full-chemistry simulation of this study. The GEOS-12 Chem tagged CO simulation includes the tracers of primary anthropogenic (fossil fuel + biofuel) 13 and BB sources, and secondary oxidations from CH4 and NMVOCs. Descriptions of all these tracers 14 are summarized in Table 2  in Fig. 2. The surface CO magnitudes and the hour-to-hour variations in Naqu, Qamdo, and Diqing 20 are higher than those in other cities in all seasons. Furthermore, the daily peak-to-trough contrast in 21 Naqu, Qamdo, and Diqing are also larger than those in other cities. The highest surface CO hourly 22 mean are typically observed in Naqu in all seasons except in the second half day (after 12:00 local 23 time (LT)) in autumn and winter (September-October-November/December-January-February 24 (SON/DJF)), when the highest surface CO values are observed in Qamdo. 25 Diurnal cycles of surface CO VMR in all cities generally show a bimodal pattern in all seasons. 26 For all cities, two diurnal maximums are generally observed during 9:00 to 11:00 LT in the daytime 27 and 21:00 to 23:00 LT in the nighttime in all seasons. The timings of the daytime diurnal maximum 28 in spring and summer (March-April-May/June-July-August (MAM/JJA)) in all cities are 1 to 2 29 hours earlier than those in SON/DJF (  40 Seasonal cycle of surface CO VMR over the HTP within the period of 2015 to 2020 are shown 41 in Fig. 3. As generally observed in most cities over the HTP, surface CO VMR showed clear seasonal 42 features: (1) high levels of surface CO VMR occur in the late autumn to spring and low levels of 1 surface CO occur in summer to early autumn; (2) the variations in the late autumn to spring are 2 larger than those in summer to early autumn; (3) seasonal cycles of surface CO VMR in most cities  3 show a bimodal pattern, i.e., a large seasonal peak occurs around November − December and a small 4 seasonal peak occurs around April − May. 5 Surface CO VMR monthly mean and month−to−month variations in Naqu, Qamdo, and Diqing 6

Seasonal cycle
are higher than those in other cities in all seasons. Furthermore, the peak−to−trough contrast in Naqu, 7 Qamdo, and Diqing were also larger than those in other cities. Surface CO VMR monthly mean 8 over the HTP varied over a large range of 206.8 ± 93.5 to 1887.1 ± 1132.0 ppbv depending on 9 season and region (Table 3), where Naqu, Qamdo, and Diqing varied over 419.0 ± 221.2 to 1887.1 10 ± 1132.0 ppbv, and other cities varied over 206.8 ± 93.5 to 759.4 ± 473.8 ppbv (Table 3). 11

Interannual variability 12
Biweekly mean time series of surface CO VMR over the HTP from 2015 to 2020 along with 13 the fitted results by using the regression model ( ) are shown in Fig. 4 (3) disclosed that large seasonal enhancements typically occur around November − 18 December and April − May which correspond to the timings of the seasonal peaks for most cities. 19 The trend in surface CO VMR from 2015 to 2020 over the HTP spanned a large range of (-21.6 ± 20 4.5) % to (11.9 ± 1.38) % per yr, indicating a regional representation of each dataset. Surface CO 21 VMR in Ngari, Lhasa, Shannan, Naqu, Qamdo, and Diqing showed negative trends. The largest 22 decreasing trends were observed in Qamdo and Naqu, which showed decreasing trends of (-16.98 23 ± 4.37) % and (-21.6 ± 4.5) % per yr, respectively. Surface CO in Shigatse and Nyingchi showed 24 positive trends. A large increasing trend of (11.9 ± 1.38) % per yr was observed in Shigatse. 25 Surface CO VMR annual mean over the HTP varied over 318.3 ± 71.6 to 901.6 ± 472.2 ppbv 26 depending on year and region ( meteorology of the HTP is not found in the literature. Here we first use IASI CO total column from 42 2015 to 2020 over the HTP to evaluate the model performance in the specifics of the HTP. As the 1 vertical resolution of GEOS-Chem is different from the IASI observation, a smoothing correction 2 was applied to the GEOS-Chem profiles. First, the GEOS-Chem CO profiles were downgraded to 3 the IASI altitude grid to ensure a common altitude grid. Since the IASI overpass time is at about 4 09:30 LT in the morning, only the GEOS-Chem simulations at 9:00 and 10:00 LT are considered. 5 The interpolated profiles were then smoothed by the monthly mean IASI averaging kernels and a 6 priori profiles (Rodgers, 2000;Rodgers and Connor, 2003). The GEOS-Chem CO total columns 7 were calculated subsequently from the smoothed profiles by using the corresponding regridded air 8 density profiles from the model. Finally, the GEOS-Chem total column time series were averaged 9 by month and compared with the IASI monthly mean data. 10 Correlation plots for the model−to−IASI data pairs in each region over the HTP are shown in 11 Fig. 5. Depending on regions, the GEOS-Chem simulations over the HTP tend to underestimate the 12 IASI observations by 9.2% to 20.0%. The largest GEOS-Chem vs. IASI differences occur in Qamdo 13 and Lhasa, with underestimations of 20.0% and 18.5%, respectively. The least GEOS-Chem vs. 14 IASI difference occurs in Nyingchi with an underestimation of 9.2%. These GEOS-Chem vs. IASI 15 differences over the HTP were mainly attributed to the underestimation of local emission inventories 16 and the coarse spatial resolution of the GEOS-Chem model grid cells. is used for investigating the influence of long range transport. We turn off all emission inventories 41 within the HTP in the GEOS-Chem tagged CO simulation and assess the relative contribution of 42 each source and geographical tracer. The relative contribution of each tracer is calculated as the ratio 43 of the corresponding absolute contribution to the modelled total concentration amount. Taking this 44 ratio effectively minimizes the propagation of systematic model errors that are common to all tracers, 1 i.e., the uncertainties in meteorological fields, the vertical mixing and STE schemes, and the 2 mismatch in spatial resolution. additional sources of CO could exist, e.g., from long range transport or oxidation from CH4 and 21 NMVOCs originating either nearby or in distant areas. 22 The emission from coal-burning for heating was thought to be the dominant sources of primary source results in the highest correlation between NO2 and CO concentrations in Lhasa. In contrast, 25 Qamdo, Naqu, and Diqing are surrounded by alpine farmlands and pastures. Historically, post-26 harvest crop residue (e.g., highland barley straws and withered grass) was often burned by local 27 farmers to fertilize the soil for next planting season. As a fine fuel, post-harvest crop residue was 28 often burned directly in the field in large piles and smolder for weeks. These seasonal crop residue 29 burning behaviors typically occur in the cold season which could cause a high level of CO emission 30 in this period. Furthermore, local residents extensively use dry yak dung as fuel for cooking or 31 heating throughout the year which could elevate the background CO level in these regions. As a 32 result, these higher local sources might be an important factor explaining the higher CO magnitude 33 in these regions. 34 35 Monthly mean contributions of anthropogenic, BB, and oxidation from long range transport to 36 the surface CO VMR over the HTP are shown in Fig.7. All statistical results are based on GEOS-37

Long range transport
Chem tagged CO simulations by turning off the emission inventories within the HTP. Due to the 38 influence of seasonally variable transport and magnitude of the regional emissions, the 39 anthropogenic, BB and oxidation sources are all seasonal and regionally dependent. Generally, 40 anthropogenic contributions in June − September and DJF are higher than those in the rest of the 41 year. In contrast, high levels of oxidation contribution occur in JJA/SON and low levels of oxidation 42 contribution occur in MAM/DJF. For BB source, contributions in MAM/DJF are larger than those 1 in JJA/SON. Depending on season and region, relative contributions of anthropogenic, BB, and 2 oxidation transported to the surface CO VMR over the HTP varied over 35 to 61%, 5 to 21%, and 3 30 to 56%, respectively. The combination of anthropogenic and oxidation sources dominated the 4 contribution which varied over 80 to 95% with an average of 89% throughout the year. 5 After normalizing each regional anthropogenic contribution to the total anthropogenic 6 contribution, the normalized relative (NR) contribution of each anthropogenic region to the total 7 anthropogenic associated transport is obtained in Fig.8. The results show that the anthropogenic 8 associated transport is mainly attributed to the influence of anthropogenic sources in South Asia and 9 East Asia (SEAS). The NR anthropogenic contribution in SEAS ranges from 58% in DJF to 91% in 10 SON. In addition, moderate anthropogenic contributions from North America (NA) (10 to 27%), 11 Europe and Boreal Asia (EUBA) (4 to 12%), and rest of world (ROW) (4 to 10%) are also observed 12 in MAM/DJF. By using a similar normalized method, the NR contributions of each BB tracer and 13 oxidation tracer are obtained in Fig.9 and Fig. 10 contribution are attributed to CH4 oxidation, and 32 to 55% of oxidation contribution are attributed 20 to NMVOCs oxidation. High-level NR contributions of CH4 oxidation occur in the cold season 21 (November -March) and low-level NR contributions of CH4 oxidation occur in the warm season 22 (April -October). The NR contributions of NMVOCs oxidation varied over an opposite mode to 23 that of CH4 oxidation; they maximize in the warm season and minimize in the cold season. The 24 JJA/SON meteorological conditions that show stronger solar radiation, higher temperature, wetter 25 atmospheric condition, and lower pressure than those in DJF/MAM are more favorable for 26 increasing VOCs emissions from biogenic sources (BVOCs), which consolidates the fact that 27 contributions of NMVOCs oxidation in warm season are larger than those in cold season. 28 By minimizing the propagation of model errors that are common to all tracers (see section 4), 29 the major factors impacting the model interpretation are the uncertainties in emission inventories 30 and OH fields. The uncertainties in CO emission inventories mainly impact primary anthropogenic 31 and BB sources, and the uncertainties in CH4 and VOCs emission inventories, and OH fields mainly 32 impact secondary oxidation sources. Additional factors that affect the generation and deplete 33 chemistry of CO or its precursors (e.g., uncertainties in emission inventories of other atmospheric 34 components, stratospheric intrusion of ozone and chemical mechanism, etc.) could also contribute 35 to the uncertainty of the interpretation. All these factors may be seasonal and regionally dependent. 36 A series of GEOS-Chem sensitivity studies might be able to quantify these uncertainties, but this is 37 beyond the scope of present work. 38 From section 5.1 and the model interpretation here, we can conclude that the main source of 39 CO in urban areas over HTP is due to local and SEAS anthropogenic and biomass burning emissions, 40 and oxidation sources. In contrast, black carbon in most of the HTP is largely attributed to Southeast 41 Asian biomass burning, and locally sourced carbonaceous matter from fossil fuel and biomass 42 combustion also substantially contribute to pollutants in urban cities and some remote regions,

Transport pathways 3
The 3D transport trajectories of CO originated in various source regions and sectors to the HTP 4 are identified as bellow. First, the GEOS-Chem tagged CO simulation is applied for determining 5 seasonal NR contribution of each tracer (Figs. 8 and 9). For the tracer with a NR contribution of 6 larger than 30% at a specific time (hereafter enhancement time), the global CO distribution provided 7 by the GEOS-Chem simulation is applied to search for potential CO sources occurring before the 8 enhancement time within 15 days. Then, we generated a series of back trajectories with various 9 travel times to judge whether these CO emissions are capable of travelling to the measurement 10 region. For instance, with respect to each CO enhancement measured at a specific time, we 11 generated ten back trajectories arriving at 100 m above the ground but with different travel time 12 ranging from 3 to 15 days. If the back trajectories intersect a region where the GEOS-Chem 13 simulation indicates an intensive CO source and the travel duration is within ± 2 hr of the observed 14 enhancement, then this specific CO source could contribute to the observed enhancement over the 15 HTP. The transport trajectories for this CO source are finally determined. Meanwhile, GEOS-Chem 16 emission inventories are used to classify this CO source into anthropogenic or BB source. This CO 17 source is regarded as BB source if GEOS-Chem BB inventory indicates an intensive CO 18 enhancement. Otherwise, it is regarded as anthropogenic source. 19 Fig. 11 demonstrates travel trajectories of polluted air masses originated in AF, SEAS + OCE, 20 EUBA, and NA regions which arrived at Naqu (31.5°N) over the HTP through long range transport. 21 As the GEOS-Chem BB inventory shown, CO emissions from southern Africa during July -22 September, central Africa during November -February, central Europe during July -November, 23 Siberia during June -September, and South Asia peninsula during March -May are dominated by 24 BB source. Other potential CO sources are dominated by anthropogenic emissions. Fig.12 shows 25 the spatial distribution of CO VMR along with the mean horizontal wind vectors at 500 hPa in 26 different seasons. Fig. 13 illustrates the latitude − height and longitude − height distributions of CO 27 VMR along with the 3D atmospheric circulation patterns in different seasons. The 3D transport 28 pathways of CO around the HTP are thus deduced as follows. 29 As indicated by the arrows in Fig. 12 and Fig.13, the strong surface cooling in DJF over the 30 HTP results in divergence and the formation of an enhanced local circulation cell, while in JJA air 31 masses converge toward the HTP from the surroundings triggered by the ascending of strongly 32 heated air masses over the HTP (Zhang et al., 2015). In DJF, the tropical easterlies are weak but the 33 mid-latitude westerlies extend to subtropics (~ 20°N) near the surface and tropics (~10°N) over 34 middle troposphere (Fig. 13). In the summer monsoon season, the atmospheric circulation patterns 35 around the HTP change dramatically and is dominated by the reversal of surface wind regime in the 36 tropics such as South China Sea, Bay of Bengal, and Arabian Sea (Fig.11 and Fig.13). Meanwhile, 37 the mid-latitude westerlies in JJA recede to the North Temperate Zone (north of 30°N) and the 38 westerly jet center shifts to about 40°N (from about 30°N in DJF). In JJA, the tropical region in the 39 south of the HTP is characterized by the strong easterlies in the upper troposphere and by the 40 southwesterly air flow in the lower troposphere (Fig. 13). The prevailing winds during the transition 41 seasons in MAM and SON are still westerlies (Fig. 13). These above seasonal atmospheric 42 circulation patterns control the CO transport pathway around the HTP. Nevertheless, the transported 43 In the SON/DJF, a significant amount of CO from southern SEAS (anthropogenic source), 3 northern AF (BB source), western EUBA, and northern NA (anthropogenic source) can be 4 transported to the HTP along the westerlies in the dry winter monsoon conditions. CO originating 5 in distant regions such as western EUBA and NA reaches a high altitude (to 8 km) during the 6 transport (Fig. 11). However, CO from the densely populated and industrialized areas in eastern 7 China barely reaches the HTP because of strong removal along the transport pathways to the HTP 8 which circles around the Northern Hemisphere along the westerlies during the winter monsoon 9 season (Fig. 12). In MAM, CO emissions from BB sources in SEAS region can be transported to 10 the HTP which is mainly triggered by deep convection followed by northward transport into the 11 mid-latitude westerlies (Liu et al., 2003) (Fig. 13). During the South Asian summer monsoon, the 12 local abundant wet precipitation can remove a large portion of SEAS originated CO but can still 13 affect southwest HTP (Fig. 13) transports local emissions far away (Fig. 13). 20

Factors driving surface CO variability over the HTP 21
Temporal CO burden is dependent on the difference between the CO source and sink, which is 22 determined by the accumulated influence of local emission, transport, secondary generation, 23 environmental capacity, and OH oxidation capability. The environmental capacity is determined by 24 atmospheric self−clean capability, topography, deposition, and meteorological condition 25 (Hofzumahaus et al., 2009). Atmospheric self-clean capability refers to the capability of the 26 atmosphere in terms of depleting atmospheric pollutants through physical and chemical processes 27 (Rohrer et al., 2014). Generally, the vertical self−clean capability is positively correlated with the 28 PBLH and the horizontal self−clean capability is positively correlated with the wind speed (Rohrer 29 et al., 2014). The OH oxidation capability is positively correlated with temperature, radiation and 30 OH seasonality (Rohrer et al., 2014). 31 The bimodal pattern of diurnal cycles for surface CO VMR in urban areas over the HTP is 32 attributed to the following diurnal production and depletion processes. Similarly, the seasonal cycle of surface CO VMR in the urban areas over the HTP is determined 43 by the seasonal variability of CO source, environmental capacity, and OH oxidation capability. High 1 levels of surface CO VMR in the late autumn to spring can be attributed to low PBLH and OH 2 oxidation capability but high local and transported CO in the period, and vice versa for low levels 3 of surface CO VMR in summer to early autumn (Yin et al., 2019b). Specifically, local anthropogenic 4 CO sources (mainly heating activities) and crop residue burning behaviors in urban regions over the 5 HTP during the colder post-monsoon and winter months are higher than those in other seasons. 6 Meanwhile, the westerlies near the surface in SON/DJF are weaker than those in MAM/JJA, which 7 facilitate the accumulation of atmospheric pollutants (Fig. 13). Furthermore, high levels of CO are 8 observed in the late autumn to spring in neighboring SEAS countries due to intensive anthropogenic 9 emissions or BB practices ( These polluted air masses can transport to the HTP region and elevate the local CO level (Fig.13). 11 Thus, apart from local anthropogenic and BB emissions, these transported sources might be an 12 important factor explaining the high CO pollution in winter. 13 Since the crop residue burning emissions result in poor air quality that threatens local terrestrial 14 ecosystems and human health, Chinese government started to ban crop residue burning over China 15 since In this study, we quantified the variability, source, and transport of CO in the urban areas over 29 the Himalayas and Tibetan Plateau (HTP) by using in situ measurement, GEOS-Chem model tagged 30 CO simulation, and the analysis of meteorological fields. Diurnal, seasonal, and interannual 31 variabilities of CO over the HTP are investigated with ~ 6 years (January 2015 to July 2020) of 32 surface CO measurements in eight cities over the HTP. Annual mean of surface CO volume mixing 33 ratio (VMR) over the HTP varied over 318.3± 71.6 to 901.6 ± 472.2 ppbv, and a large seasonal cycle 34 was observed with high levels of CO VMR in the late autumn to spring and low levels of VMR in 35 summer to early autumn. Surface CO VMR burdens and variations in Naqu, Qamdo, and Diqing 36 are higher than those in other cities in all seasons. The diurnal cycle is characterized by a bimodal 37 pattern with two maximums occurring around 9:00 to 11:00 local time (LT) in the daytime and 21:00 38 to 23:00 LT in the nighttime. The trend in surface CO VMR from 2015 to 2020 over the HTP 39 spanned a large range of (-21.6 ± 4.5) % to (11.9 ± 1.38) % per yr, indicating a regional 40 representation of each dataset. However, surface CO VMR from 2015 to 2020 in most cities over 41 the HTP showed negative trends. 42 The IASI satellite observations are for the first time used to assess the performance of GEOS-43 Chem full-chemistry model for the specifics of topography and meteorology over the HTP. 1 Depending on the region, the GEOS-Chem simulations over the HTP tend to underestimate the IASI 2 observations by 9.2% to 20.0%. Though not perfect in reproducing the absolute values of the IASI 3 observation, GEOS-Chem can capture the measured seasonal cycle of CO total column over the 4 HTP with a correlation coefficient (r) of 0.64 to 0.82 depending on regions. Distinct dependencies 5 of CO on a short life time species of NO2 in almost all cities over the HTP were observed, implying 6 local emissions to be predominant. By turning off the emission inventories within the HTP in GEOS-7 Chem tagged CO simulation, the relative contribution of long range transport was evaluated. The 8 results disclosed that transport ratios of primary anthropogenic source, primary biomass burning 9 (BB) source, and secondary oxidation source to the surface CO VMR over the HTP varied over 35 10 to 61%, 5 to 21%, and 30 to 56%, respectively. The anthropogenic contribution is dominated by the 11 South Asia and East Asia (SEAS) region throughout the year (58% to 91%). The BB contribution 12 is dominated by the SEAS region in spring (25 to 80%) and the Africa (AF) region in July - This study concluded that the main source of CO in urban areas over HTP is due to local and 21 SEAS anthropogenic and BB emissions, and oxidation sources. In contrast, black carbon in most of 22 the HTP is largely attributed to Southeast Asian biomass burning, and locally sourced carbonaceous 23 matter from fossil fuel and biomass combustion also substantially contribute to pollutants in urban 24 cities and some remote regions, respectively. This study not only emphasized the different origins 25 of diverse atmospheric pollutants in the HTP, but also improved knowledge of the variabilities, 26 sources, drivers, and transport pathways of atmospheric pollutants over the HTP and provided 27 guidance for potential regulatory and control actions.             Table 2 for description of each tracer. 4 5 Fig. 9. Monthly mean contributions of each geographical BB tracer to the total BB associated CO transport to the 6 Himalayas and Tibetan Plateau (HTP). Vertical error bar represent 1σ standard variation within that month. See 7 Table 2 for description of each tracer.  Table 2 for description of each tracer.  Tables 1 Table 1. Geolocations of measurement sites in eight cities over the HTP region. All sites are organised as a function 2 of increasing longitude. Population statistics are prescribed from the 2018 demographic data provided by National  Table 3. Statistical summary of surface CO VMR in eight cities over the HTP region. All cities are organised as a