2.1 Site description

The field study was conducted between 31 May and 1 July 2015 at a high-altitude observatory, i.e., Nam Co station for Multisphere Observation and Research, Chinese Academy of Sciences (90^∘57^′ E, 30^∘46^′ N; 4730 m a.s.l.), at the central part of the TP (Fig. 1). The Nam Co station is located near Nam Co Lake (Fig. 1b), the second largest inland lake in the TP (area: 1920 km²) which is located at the northern border of the Nyainqentanglha Mountains. The melted glacier from the Nyainqentanglha Mountains supplies water to the lake each year during the warm season. The Nam Co station and its surrounding is a pristine region except for a small village that is about 10 km west of the station with a population of 300 to 500. In the past several years, tourism at this beautiful lake has grown. A highway for tourists was built about 3 km south of the station with low traffic intensity (less than 300 cars per day) during June. The capital city (Lhasa) of the Tibet Autonomous Region is about 100 km southeast of the station, with an average elevation of 3600 m a.s.l., between which are the Nyainqentanglha Mountains (higher than 6000 m a.s.l.; Fig. 1c). The closest town, Dangxiong, is alongside the famous Qinghai–Tibet Highway and about 70 km east of the station, with an average elevation of 4200 m a.s.l. Overall, the station is surrounded by the mountains in the south and east, and the lake to the west. The ecology of surrounding area is semi-arid land dominated by alpine meadow and barren areas. The precipitation mainly occurs during summer monsoon period. The cooking and heating at the station uses electric power and natural gas.

2.2 Instrument setup

The study was conducted at the observatory field of the station using a custom-made trailer with inlet stepped out of the top with the height of ∼ 5 m above ground. All the instruments were arranged inside the trailer where the air temperature was controlled at ∼ 20 ^∘C by two air conditioners. The air flow of inlet was maintained by a vacuum pump with a flow rate of 10 L min⁻¹, and several instruments took samples from this inlet including a high-resolution time-of-flight aerosol mass spectrometer (HR-AMS, Aerodyne Research Inc., Billerica, MA, USA), a custom-made scanning mobility particle sizer (SMPS), a soot particle Aerodyne mass spectrometer (SP-AMS, Aerodyne Research Inc., Billerica, MA, USA), a single-particle intracavity laser-induced incandescence photometer (SP2, DMT, Inc., Boulder, CO, USA), and a seven-wavelength aethalometer (model AE31, Magee Scientific, Berkeley, CA, USA). The total flow rate of the inlet was maintained at ∼ 16 L min⁻¹ and a PM_2.5 cyclone was used in the front of the inlet (model URG-2000-30EH, URG Corp., Chapel Hill, NC, USA) to remove coarse particles. Several additional instruments were also co-located but with their own inlets, such as a nephelometer (model 3563, TSI, Shoreview, MN, USA), a PM_2.5 sampler (model PQ200, BGI, USA), and a multi-angle absorption photometer (MAAP, model 5012, Thermo Electron Cooperation, USA). The meteorological data were recorded at the tower of the Nam Co station at a height of 20 m above ground.

HR-AMS was used to measure the non-refractory PM₁ (NR-PM₁), which was thermally vaporized at ∼ 600 ^∘C, ionized with a 70 eV electron impact and determined by using a time-of-flight mass spectrometry. The details of the instrument have been described elsewhere (DeCarlo et al., 2006). For our study, HR-AMS was only operated with V mode with 5 min resolution due to the low aerosol mass loading. Due to malfunction of the chopper, the size distribution of NR-PM₁ was not determined. The HR-AMS was calibrated for ionization efficiency (IE) with the ammonium nitrate following standard procedures (Jayne et al., 2000) at the beginning, middle, and end of the study. Particulate-free air was sampled twice during the study period to adjust the influence of air on the fragmentation table and determine the detection limits (DLs) of aerosol species. The DLs were calculated as 3 times the standard deviations (3σ) of the measured values during the particle-free sampling periods. The 2.5 min DLs for organic, sulfate, nitrate, ammonium, and chloride were determined at 0.108, 0.014, 0.007, 0.002, and 0.010 µg m⁻³, which are comparable to the values reported in previous studies.

Table 1Summary of AMS measurement at high-elevation sites around the world. The mass concentration was for NR-PM₁.

Download Print Version | Download XLSX

2.3 Data processing

The HR-ToF-AMS data were processed using the standard software of SQUIRREL (v1.56) and PIKA (v1.15c) written in IGOR (Wavemetrics, Inc., Lake Oswego, OR, USA; http://cires.colorado.edu/jimenez-group/ToFAMSResources/ToFSoftware/index.html) to determine the mass concentrations of the NR-PM₁ species and the ion-speciated mass spectra of organics. An empirical particle collection efficiency (CE) of 0.5 was used, which has been widely used in field studies employing AMS with a dryer installed in front of the equipment's particle inlet. This CE value was further validated by the consistency and reasonable slope between HR-AMS-measured mass concentrations and SMPS-determined particle volumes (Sect. 3.1.2, R²= 0.9, slope = 1.48). The elemental ratios of OA (O ∕ C; H ∕ C; and organic matter to carbon, OM ∕ OC) in this study was determined using the “improved-ambient” method (Canagaratna et al., 2015). Default relative ionization efficiency (RIE) values were used for organics (1.4), nitrate (1.1), and chloride (1.3), while an RIE value of 3.8 was determined for ammonium and 1.1 for sulfate based on the calibration for pure NH₄NO₃ and (NH₄)₂SO₄, respectively.

Figure 2The combo plot of the data of the Nam Co study including (a) the meteorological conditions (T: air temperature; RH: relative humidity; Precip.: precipitation), (b) the variation of WS (wind speed) colored according to WD (wind direction), (c) the temporal variation of mass concentration of PM₁ species and the average contribution each species (pie chart), (d) the mass contribution of each PM₁ species and the total mass concentration of PM₁, and (e) the mass contribution of PMF results (Sect. 3.5). Three periods based on the meteorological conditions were marked. Please note the dates are given in MM/DD format.

Download

The source apportionment of organics was conducted by PMF with the robust engine. First, organic matrix was analyzed using the PMF2.exe algorithm in robust mode (Paatero and Tapper, 1994) and explored using the PMF Evaluation Toolkit (PET; Ulbrich et al., 2009). The PMF solution was evaluated following the procedures outlined in Table 1 of Zhang et al. (2011) including modification of the error matrix and down-weighting of low signal-to-noise ions. Moreover, based on the AMS fragmentation table, some organic ions were not directly measured but scaled to the organic signal at m∕z 44, which were down-weighted by increasing their errors by a factor of 3. A two-factor solution with fPeak = 0 was chosen in this study, as it is able to reconstruct the total OA mass and temporal profiles very well. The results of three-factor solution with fPeak = 0 are shown in the Supplement (Fig. S1), which show splitting of the solutions.

3.1 The meteorological conditions during the study

During the field study, the meteorological conditions were generally cold and windy (Fig. 2a and b). The average air temperature was 8.4 ± 3.6 ^∘C with a diurnal hourly average ranging from 4.8 to 12.3 ^∘C; the diurnal average wind speed (WS) ranged from 4.5 to 6.5 m s⁻¹. The dominant wind directions were south and southwest, although they varied during different periods (Figs. 2b and S2). The wind direction (WD) had distinct diurnal variation with air mass from south during night to morning and west during afternoon. Based on the report of the Climate Diagnostics Bulletin of India (http://imdpune.gov.in/Clim_RCC_LRF/Climate_Diagnostic_Bulletins/cdbi_apr_2015.pdf), the precipitation from Indian monsoon extended over most of India over 12 to 14 June and these days are treated as the onset of South Asian monsoon. Indeed, precipitation was recorded at the Nam Co station on 13 June and lasted for several days (Fig. 2a). Based on these weather conditions, the period before 14 June was classified as the pre-monsoon period and afterwards as the monsoon period. The RH during the pre-monsoon and monsoon periods were 48.4 ± 19.2 and 58.8 ± 16.5 %, respectively. The origins of the air masses were also different during these two periods. Figure 3 shows the air mass from west during pre-monsoon period accounted about 30–40 %, while it was only 7 % during monsoon period, of which ∼ 80 % was from south.

3.2 The temporal variations of chemical species

The average mass concentration of PM₁ (NR-PM₁+ BC) was 2.0 µg m⁻³ during the whole study period with 68 % OA, 15 % sulfate, 2 % nitrate, 7 % ammonium, and 8 % BC. The NR-PM₁ mass loading (1.84 µg m⁻³) at Nam Co station was lower than the values observed at some high-elevation sites such as Menyuan (10.8 µg m⁻³; Du et al., 2015), Montsec (4.9 µg m⁻³; Ripoll et al., 2015), Mt. Cimone (4.5 µg m⁻³; Rinaldi et al., 2015), Puy de Dôme (5–27 µg m⁻³), and Mt. Bachelor (15.1 µg m⁻³; Zhou et al., 2017), but higher than those in some other sites, such as sub-Antarctic (0.46 µg m⁻³; Schmale et al., 2013) and Jungfraujoch (0.55 µg m⁻³; Fröhlich et al., 2015), similar to that at Mt. Whistler (1.91 µg m⁻³; Sun et al., 2009; Table 1 and Fig. S3). Although the sampling seasons and altitudes of these sampling sites are different (Table 1), the results of these studies could basically represent the level of aerosol at these sites due to their relatively long sampling spans (from 1 to 10 months). The contribution of OA at these high-elevation sites ranged from 50 to 90 %; the highest value was at Mt. Bachelor, which was frequently influenced by transported biomass burning plume, while the relative low OA contribution (38 %) in the sub-Antarctic due to the dominant source from marine emissions with a high sulfate content. The mass concentration of PM₁ varied dynamically during our study with distinct difference between pre-monsoon and monsoon periods. The average mass concentrations of PM₁ for these two periods were 2.6 and 1.2 µg m⁻³, respectively. For comparison, the concentrations of various species for these two periods were displayed side by side in Fig. 4a. The ratios between pre-monsoon and monsoon for all the species were higher than 1 with the maxima for ammonium (3.1) and sulfate (2.8); the contribution of OA was thus slightly higher during the monsoon period than during the pre-monsoon (71 vs. 64 % with a mass concentration ratio of 1.9). Apart from the potential scavenging effect, these results could also be influenced by the sources, transport route, and chemical processes during different periods.

Figure 372 h backward air mass trajectory and class statistics over Nam Co station for (a) P1, (b) P2, and (c) monsoon period. The back trajectories at 1000 m above ground level were calculated at 1h intervals using the NOAA Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT) model (Draxler and Hess, 1998) and 1^∘ resolution Global Data Assimilation System (GDAS) data set (https://ready.arl.noaa.gov/gdas1.php). The classes of trajectories are colored according to pressure. Fire spot observed by MODIS (https://firms.modaps.eosdis.nasa.gov) and average wind rose plot during each period are also shown.

Download

Based on the mass concentration and temporal variations of PM₁ species and weather conditions, the pre-monsoon period could be further divided into two periods, i.e., period 1 (P1: 30 May to 7 June) and period 2 (P2: 8 to 13 June). P1 was characterized by relatively high sulfate concentration (0.46 vs. 0.41 µg m⁻³) and sunny days, while P2 was characterized by high PM₁ concentration (2.9 vs. 2.1 µg m⁻³; P < 0.005, χ² test), high nitrate contribution, and wet and cold days (Fig. S4). The air masses for P2 had higher contributions from west and north than those during P1 (Fig. 3). Figure 4b shows the comparisons of mass concentration of different species between these two periods. For chemical species, from P1 to P2, nitrate and OA increased dramatically by a factor of 1.6 and 2.2, respectively, and BC and ammonium also increased, suggesting the influence of the elevated air plume. However, sulfate and chloride decreased slightly (the ratios between P2 and P1 were ∼ 0.95). The variations of sulfate and nitrate during these two periods could be related to the photochemical conditions and origination of air mass (see Sect. 3.3).

Figure 4The comparisons of PM₁ species of the mass concentration (left axis) and the ratio between them (right axis) between (a) pre-monsoon and monsoon and (b) P1 and P2.

Download

The particles were generally neutralized as illustrated by the scatter plot between predicted and measured ammonium (slope = 0.91; Fig. S5a). The neutralized PM₁ were likely due to the high availability of ammonia from agriculture emission in northern India (Clarisse et al., 2009; Van Damme et al., 2015). The slope was ∼ 0.75 during P2 (green circles, Fig. S5a), which suggests that there was more ammonium than needed to neutralize sulfate and nitrate. We checked whether the high ammonium was from fragmentation of organic nitrogen such as amines. The ratios of NH⁺ to NH${}_{\mathrm{2}}^{+}$ were consistent during the study (Fig. S5b), suggesting there was no significant influence from amino compounds. This excess ammonium determined by the neutralization could be related to the presence of significant amounts of organic anions in aerosol (such as carboxylic acids) or variation of RIE for ammonium, which could have a higher value in the mixed acidic particles.

Figure 5Diurnal variations of each species and weather conditions (RH: relative humidity; Tem: air temperature; and WS: wind speed) during three periods of the study. Note that the signals of sulfate, ammonium, and BC are increased by a factor of 2 or 1.5 for comparison, respectively.

Download

3.3 Diurnal variation and chemical process of different species

The diurnal cycles of OA, sulfate, nitrate, ammonium, and BC during different periods are shown in Fig. 5. All these species unexpectedly present dramatic diurnal variations, especially during P2. OA, sulfate, ammonium, and BC showed a similar pattern with low values during nighttime to early morning and high values during afternoon, which suggests their common sources or similar transport pathways (Zhang et al., 2001). We checked the diurnal variation of the origination of air masses and found that there were increased air masses from south during nighttime and from west during afternoon (Fig. S6). This phenomenon could be related to the plateau monsoon during summer, when the strong heating at the surface resulted in a shallow cyclonic circulation over the TP (Tang and Reiter, 1984). The enhanced air mass transportation from west during afternoon could favor the aerosol plume transport. The enhanced WS during afternoon was also observed with the increase in air temperature (Fig. 5). The diurnal variations of chemical species during monsoon period were relatively flat compared with those during P1 and P2, which may be related to the relatively consistent air mass origination during monsoon. Nitrate presented a significantly different diurnal variation with high values during nighttime to early morning and low values during afternoon. These features were highly correlated with that of the RH and air temperature (Fig. 5), suggesting the importance of thermodynamically driven gas/particle partitioning of ammonium nitrate and heterogeneous production of nitrate due to hydrolysis of N₂O₅. In addition, during the early morning (06:00–08:00), there was a peak for most species, which was accompanied by the lowest air temperature and the highest RH, and the lower plenary boundary layer (PBL), which could concentrate all the aerosol species (Yanai and Li, 1994). Overall, the diurnal variations of aerosol species at Nam Co may be dominated by the variabilities of both long-range transport air mass and local meteorological conditions.

Figure 6Bivariate polar plots that illustrate the variations of the concentrations (colored) of each species as a function of wind speed (m s⁻¹) and wind direction during different periods of the study.

Download

In order to further elucidate the chemical processes and potential sources of the aerosol species, the relationships of chemical species with wind conditions were analyzed based on bivariate polar plot analysis (Fig. 6). During P1, sulfate and ammonium had hotspots from all directions across a wide range of wind speed (0–10 m s⁻¹); OA was prevailing from southeast and northwest winds; nitrate had hotspots when the wind speed was relatively slow (0–8 m s⁻¹) and from southeast/east mainly; BC had hotspots from south, west, and northwest at high wind speeds (4–12 m s⁻¹). During P2, all species except nitrate had hotspots from south/southwest at high wind speeds (6–14 m s⁻¹); nitrate also had hotspots at a relative low wind speed from southwest. These results suggested that the sources and formations of aerosols during P1 and P2 could be different, although the diurnal variations of species between these two periods were similar. During monsoon period, all species had similar hotspot distributions, which were from all directions but relatively weak from southeast.

The significant difference of diurnal variation between sulfate and nitrate suggested the different chemical evolution of these two species. Nitrate and sulfate during three periods all had weak correlations. During P1, the ratios of sulfate to nitrate had peaks (> 20) during afternoon, while during P2, all ratios remained low (< 20; not shown). In order to further investigate the chemical evolutions of these two species, we selected the high-concentration periods based on nitrate concentration (> average +2σ). Figure 7 showed all periods with high nitrate during the study and the corresponding meteorological conditions. It is easy to find that most of the high mass loading periods occurred during nighttime. The wind speed and wind direction varied dynamically and mostly were from southwest. Higher wind speed from this wind direction could transport more aerosol plumes to Nam Co as illustrated from event 1 (E1) during which all the species (OA, sulfate, ammonium, and BC) except nitrate increased; this type of event was accompanied by warm and dry air conditions. When the wind direction was from southwest with lower wind speed (E2), the RH increased to higher than 90 % accompanied by deceased air temperature, and the mass concentrations of nitrate, sulfate, and OA increased significantly and BC decreased significantly. These results further suggest the different chemical evolution and sources for different aerosol episodes at Nam Co station.

Figure 7The high aerosol loading periods based on nitrate (> average + 2σ) accompanying meteorological data.

Download

3.4 The average chemical feature of organic aerosol

The average mass spectrum of OA was shown in Fig. 8a. The organic mass was on average composed of 51 % oxygen, 44 % carbon, and 5 % hydrogen, with an average nominal formula being C₁H_1.33O_0.88N_0.001. On average, C_xH_yO${}_{\mathrm{1}}^{+}$ (38.1 %) and C_xH${}_y^{+}$ (37.3 %) ions dominated the total OA followed by C_xH_yO${}_{\mathrm{2}}^{+}$ (19.8 %), and H_yO${}_{\mathrm{1}}^{+}$ (4.6 %). m∕z 44 was the base peak in the OA spectrum and mainly composed of CO${}_{\mathrm{2}}^{+}$ (99 %). m∕z 43 had contributions from both C₂H₃O⁺ (82 %) and C₃H${}_{\mathrm{7}}^{+}$ (18 %). m∕z 55, likely an important primary fragment, consisted of 51 % C₄H${}_{\mathrm{7}}^{+}$ and 49 % C₃H₃O⁺. The OA was highly oxidized with an O ∕ C ratio of 0.88 on average. The OA was more oxidized during pre-monsoon than monsoon with higher O ∕ C ratio (0.94 vs. 0.72) and lower H ∕ C ratio (1.28 vs. 1.44; Fig. 8b). The oxidation states of OA during two pre-monsoon periods were also different. The O ∕ C during P2 (0.98) was higher than P1 (0.90) and the H ∕ C was higher during P1 than P2 (1.30 vs. 1.27, significant, P < 0.005, χ² test). Correspondingly, the OA during P2 contained higher contribution of C_xH_yO${}_{\mathrm{1}}^{+}$ (40.3 vs. 39.1 %, significant, P < 0.005, χ² test) and C_xH_yO${}_{\mathrm{2}}^{+}$ (25.2 vs. 23.6 %, significant, P < 0.005, χ² test) ions than those during P1 (Fig. 8c).

Figure 8(a) The average mass spectrum of organic aerosol, (b) the average ratios of H ∕ C and O ∕ C during different periods, and (c) the average contribution of six ion categories during different periods.

Download

The diurnal variations of both O ∕ C and OM ∕ OC ratios presented higher values during late morning to afternoon and lower values during early morning (Fig. 9a). H ∕ C presented an opposite trend. These patterns suggested that OA was more oxidized during daytime which could be due to photo oxidation and/or transport of highly oxidized OA during daytime. We examined the variation of elemental ratios with OA mass concentrations, and found that the O ∕ C increase was accompanied by the increase in OA mass concentration (Fig. 9b). This relationship could be related with the different influences of wet scavenging on more oxidized and less oxidized OA. In addition, this result suggested the likely importance of transportation on the oxidized OA during afternoon due to the higher mass concentration occurring frequently in the afternoon. However it was not possible that aerosol plume arrived at Nam Co within several hours due to the long distance from source regions (more than 1000 km from India). Previous studies have reported the presence of an aerosol layer between 6 and 18 km a.s.l. altitude over the Tibetan Plateau during summer (Tobo et al., 2007; Vernier et al., 2011). He et al. (2014) examined the vertical profiles of aerosol extinction coefficients measured with a micro-pulse lidar at Naqu, about 100 km east of Nam Co station, and observed a maximum in aerosol extinction coefficient between 18 and 19 km a.s.l. during summer 2011. Recently, Gu et al. (2016) examined the aerosol compositions using the global three-dimensional Goddard Earth Observing System chemical transport model (GEOS-Chem) and found elevated concentrations of sulfate, nitrate, ammonium, BC, and organic carbon over the TP. Further, observational and modeling studies have also shown that deep convection over the TP during daytime is one of the important routes for tropospheric and stratospheric exchange of aerosols (Cristofanelli et al., 2009, 2010; Lin et al., 2016). Thus the enhanced aerosol concentrations during afternoon could possibly be attributed to the downward mixing of the aerosol layer from the upper troposphere during the growth of TP boundary layer. Recently, Xu et al. (2017) elucidated this mechanism at Nam Co station to explain the source of O₃ and peroxyacetyl (PAN), and found strong downward motion core events during pre-monsoon and monsoon periods which were accompanied by increased O₃ and PAN at the ground level. We also performed a test using the Weather Research and Forecasting (WRF) model to examine this kind of transport during P2 which indeed showed a downward motion core at the Nam Co station (Fig. S7). Nevertheless, this hypothesis needs further validation in the future in this region.

Figure 9(a) Diurnal variations of elemental ratios and (b) the variations of elemental ratios as the function of mass concentration of organic aerosol.

Download

3.5 The PMF apportionment on the OA

The two OA factors determined by PMF have distinctly different mass spectra and diurnal patterns (Fig. 10), including a less oxidized oxygenated organic aerosol (LO-OOA; O ∕ C = 0.49) and a more oxidized oxygenated organic aerosol (MO-OOA; O ∕ C = 0.96). Both factors appeared to be secondary in nature. Our inability to separate a primary traffic-related OA factor is consistent with the fact that C₄H${}_{\mathrm{9}}^{+}$ was a minor peak in the OA spectra (0.6 % of the total signal) and its organic-equivalent concentration averaged only 0.008 µg m⁻³ during this study, suggesting a minimal contribution from traffic-related primary OA. Collier et al. (2015) reported that the average fC₄H${}_{\mathrm{9}}^{+}$ (fraction of total organic signal accounted for by C₄H${}_{\mathrm{9}}^{+})$ was 8.1 % in the mass spectrum of primary OA from vehicle emissions. Based on this relationship, even assuming C₄H${}_{\mathrm{9}}^{+}$ was completely contributed by vehicle-derived OA, the average vehicle OA concentration would be only ∼ 0.1 µg m⁻³, or 7 % of total OA mass, during our study. Note that C₄H${}_{\mathrm{9}}^{+}$ fit well in the PMF analysis with very small residual, indicating that the signal of this ion was properly apportioned between the two OOA factors.

The mass spectra of MO-OOA and LO-OOA were characterized by high peaks at m∕z 44 (mostly CO${}_{\mathrm{2}}^{+})$ and LO-OOA had a relative large peak at m∕z 43 (mostly C₂H₃O⁺) as well (Fig. 10a). The contributions of the C_xH_yO${}_{\mathrm{2}}^{+}$ ion category in these two mass spectra were 15.1 and 28.6 %, respectively, while the contributions of C_xH_yO${}_{\mathrm{1}}^{+}$ were 37.8 and 41.5 %, respectively. The time series of LO-OOA correlated well with C_xH_yO${}_{\mathrm{1}}^{+}$ and C_xH${}_y^{+}$ ions, while MO-OOA correlated well with C_xH_yO${}_{\mathrm{2}}^{+}$ ions (Fig. S8). In addition, the time series of MO-OOA correlated well with sulfate (R²= 0.55) and BC (R²= 0.54), and less well with nitrate (R²= 0.33), while LO-OOA correlated poorly with sulfate, BC, and nitrate (Figs. 10b and S9). These results highlight the oxidation degree of MO-OOA and LO-OOA. LO-OOA and MO-OOA accounted for 41 and 59 % on average, respectively, of total OA mass during this study (Fig. 11a), and their relative contributions varied across the study (Fig. 2e). For example, LO-OOA accounted for 41 % of the total OA mass during P1, 24 % during P2, and 67 % during the monsoon period (Fig. 11b). The diurnal pattern of MO-OOA was characterized by higher concentrations during afternoon similar to those of sulfate and BC, while the diurnal pattern of LO-OOA peaked at early evening (Fig. 10c). The polar plot showed concentrated hotspots to the northwest of the sampling site for MO-OOA, and southeast for LO-OOA (Fig. 6), indicating that the sources of these two components were different. Based on trajectory analysis (Fig. 3), MO-OOA was likely closely related to long-range transport of air mass from southwest, while LO-OOA could be derived from relatively shorter distance transport such as from a marine air mass from south and regional background aerosol during the nighttime. Shen et al. (2015) reported that there was a significant aerosol source from biogenic emissions during summer near the Nam Co station. As shown in Fig. 12a and b, MO-OOA, which was highly oxidized (O ∕ C = 0.96), appeared on the upper-left corner of the triangle plot while LO-OOA was in the middle part with an O ∕ C ratio of 0.49. The high oxidation degree of MO-OOA was likely related to extensive aging processes occurring during long-range transport. The slope of linear fitting of all the data points in the VK diagram is −0.76, suggesting the evolution of OA as carboxylic acid functionalization (Fig. 12c).

Figure 10PMF results of (a) high-resolution mass spectra colored according to six ion categories for LO- and MO-OOA, respectively, (b) temporal variations of the OA factors and corresponding comparison with tracer species (sulfate and nitrate), and (c) diurnal variations of the mass concentration of the OA factors and tracer species.

Download

Biomass burning emission is an important source in South Asia and could be transported to Himalayas and TP during pre-monsoon (Engling et al., 2011; Kumar et al., 2011; Sang et al., 2013; Cong et al., 2015); however, there are not significant signals of biomass burning aerosol in our AMS results such as signals at m∕z 60 and 73 in mass spectrum of OA which were found to be associated with levoglucosan formed from the pyrolysis of cellulose (Alfarra et al., 2007). The contribution of f60 (fraction of total organic signal accounted for by m∕z 60) for LO-OOA and MO-OOA were 0.2 and 0.3 %, respectively, which were similar to the global background level (less than 0.3 %) suggested by Cubison et al. (2011). These results suggest that OA, if ever partly originating from biomass burning emissions, could have been highly oxidized during transport. The behavior that levoglucosan could be quickly (within a few hours) oxidized after being emitted has been observed in a few studies (Ortega et al., 2013). In addition, Zhou et al. (2017) recently reported the observation of a highly aged biomass burning OA (BBOA) factor with f60 < 0.3 % in its mass spectrum, in aged wildfire plumes that had gone through extensive photochemical oxidation.

Figure 11(a) The average mass contributions of two factors (a) during the study, (b) during the different periods, and (c) the diurnal mass contribution of two factors (left axis) accompanying the total organics (right axis).

Download

Figure 12Scatter plots of (a) f44 vs. f43 (dates given in MM/DD/YYYY format), (b) fCO${}_{\mathrm{2}}^{+}$ vs. fC₂H₃O⁺, and (c) H ∕ C vs. O ∕ C for the OA. The dashed lines in (a) and (b) refer to a triangular region that encompasses ambient OOA factors determined from PMF analyses of 43 AMS data sets (Ng et al., 2010). The light solid lines with slopes of 0, −1, and −2 in (c) indicate the changes of H ∕ C against O ∕ C due to adding specific functional groups to an aliphatic carbon (Heald et al., 2010); the light dashed lines indicate the oxidation state (OS) of −2, −1, 0, 1, and 2, respectively (Kroll et al., 2011); the red and blue dashed lines are derived from the right and left lines in the triangle plot; the heavy solid line indicates the slope of our measured data. The corresponding values of the OOAs identified in this study are also shown in each plot.

Download

3.6 Sensitivity of the aerosol transport to synoptic process

The synoptic process is an important factor determining if the regional emissions can be transported to the TP. It is interesting to know what kind of synoptic process is favorable for transporting aerosol plume to the Himalayas and the TP. A 3-D regional chemical transport model (REAM) coupled with WRF was used to examine the chemical evolution and regional transport of pollutants such as aromatics in this study. REAM has been used in previous studies of the Tibetan Plateau, and details about the model can be found in Zhang et al. (2017) and Supplement. REAM could capture some synoptic processes which cannot be simulated by the normally used reanalysis data due to their low resolution and the complexity of terrain in the Tibetan Plateau (Zhang et al., 2017). Figure 13 shows the distribution of simulated daily surface wind, 300 hPa geopotential height fields, and concentrations of reactive aromatics over the Tibetan Plateau during 30 May–13 June 2015. During 30 May to 7 June, corresponding to P1, there was a trough over the north propagated from west to east and this low pressure induced increasingly stronger surface wind from India to the TP, which could lead to transport of aerosol plume as illustrated by the results of model and AMS. The simulated concentration of reactive aromatics showed a peak during this period (Fig. S10). During 8–11 June, there was a weak ridge system over the north. Weak wind from west and south was observed and the simulated concentrations of reactive aromatics were sharply decreased (Fig. S10). After that, a weak low-pressure trough system was observed again. The increased concentrations of reactive aromatics were also observed accompanying the intensified southern wind. Although these trends are basically consistent with our AMS results, there were also significant differences (Fig. S10). The possible reason was that the weak ridge during P2 enhanced the wind from north and weakened the wind from west and south, where a lot of biomass burning emission sources are located (Fig. 3b). Zhang et al. (2017) suggested that a cut-off low system from the upper troposphere could be an important driver for aerosol plume transport into the TP. In our study, the trough/ridge system seems to be also an important factor affecting the transport of aerosol plume from south and west, although this effect tends to be weaker in summer than in the other seasons because the tropopause is higher and stratospheric wave activity is weaker in summer.

Figure 13Daily distribution of WRF-simulated surface wind and REAM-simulated concentrations of reactive aromatics over the Tibetan Plateau during 30 May to 13 June 2015.

Download

3.7 Atmospheric implications

Our results have several potential implications for atmospheric studies in the TP and Himalayas. Firstly, it is useful for the accurate estimation of the radiative forcing of aerosols in this region and validation of current model simulation results based on our observed chemical composition and mass loadings of fine aerosols. Ji et al. (2015) estimate the radiative forcing from aerosols over the TP and Himalayas at the surface level using a regional climate model (RegCM4.3); for carbonaceous aerosols, there are several studies which have tested the model results, but all other species referred to in published studies have coarse spatial and temporal resolution. Secondly, our findings have implications for aerosol deposition on the glacier of this region and evaluation of subsequent impacts on the melting of snow/ice (Yasunari et al., 2010). Thirdly, highly time-resolved aerosol data are very scarce in this remote plateau; thus our data are valuable to validate modeling results regarding the transport of aerosol plume as demonstrated in Sect. 3.6. Finally, the transport mechanism of aerosol to the inland of TP is less understood so far. Hindman and Upadhyay (2002) suggested that the vertical lifting due to convection and subsequent horizontal mountain valley wind could lead to the transport of aerosol from Nepal to Tibet. Dumka et al. (2010) also highlighted the important role of mountain valley wind in the aerosol transport in the central Himalayas. The dynamic variations of aerosol chemical species measured here are likely helpful for elucidating the transport mechanism of high-concentration air mass. Nevertheless, this scientific issue requires further detailed investigations in the future.