In-depth study of the formation processes of single atmospheric particles in the south-eastern margin of the Tibetan Plateau

The unique geographical location of the Tibetan Plateau (TP) plays an important role in regulating global climate change, but the impacts of the chemical components and atmospheric processing on the size distribution and mixing state of individual particles are rarely explored in the south-eastern margin of the TP, which is a transport channel for pollutants from Southeast Asia to the TP during the pre-monsoon season. Thus a single-particle aerosol mass spectrometer (SPAMS) was deployed to investigate how the local emissions of chemical composition interact with the transporting particles and assess the mixing state of different particle types and secondary formation in this study. The TP particles were classified into six distinct types, mainly including the largest fraction of the potassium-rich (K-rich) type in the total particles (30.9 %), followed by the biomass burning (BB) type (18.7 %). Most particle types were mainly transported from the sampling site's surroundings and along the Sino-Myanmar border, but the air mass trajectories from north-eastern India and Myanmar show a greater impact on the number fraction of the BB (31.7 %) and dust (18.2 %) types, respectively. Then, the two episodes with high particle concentrations showed that the differences in the meteorological conditions in the same trajectory clusters could cause significant changes in chemical components, especially the dust and aged elemental carbon (aged EC) types, which changed by a total of 93.6 % and 72.0 %, respectively. Ammonium and dust particles distribute at a relatively larger size (∼600 nm), but the size peak of other types is present at ∼440 nm. Compared with the abundant sulfate (⁹⁷HSO${}_{\mathrm{4}}^{-}$), the low nitrate (⁶²NO${}_{\mathrm{3}}^{-}$) internally mixed in TP particles is mainly due to the fact that nitrate is more volatilized during the transport process. The formation mechanism of secondary speciation demonstrates that the formation capacity of atmospheric oxidation is presumably affected by the convective transmission and the regional transport in the TP. However, the relative humidity (RH) could significantly promote the formation of secondary species, especially ⁹⁷HSO${}_{\mathrm{4}}^{-}$ and ¹⁸NH${}_{\mathrm{4}}^{+}$. This study provides new insights into the particle composition and size, mixing state, and ageing mechanism in high time resolution over the TP region.

1 Introduction

Atmospheric aerosols have complex components and sources and can be coated with inorganic or organic materials during transport and atmospheric processing (Crippa et al., 2013), and then its sizes, chemical compositions, mixing states, and optical properties would change greatly, leading to its influence in the atmosphere being more uncertain (Jacobson, 2002; Zaveri et al., 2010; Matsui, 2016; Budisulistiorini et al., 2017; Ma et al., 2012). Currently, the influences of the complex chemical components on aerosol size and mixing state show large regional differences due to the variations in the pollution sources, atmospheric formation mechanism, and meteorological conditions, which have been widely studied in an urban area at a low altitude (Pratt et al., 2011; Q. Liu et al., 2020; Xu et al., 2017; Wang et al., 2022). However, Q. Liu et al. (2020) have found that the migration or formation of low-volatility components (such as nitrate and organic matter) could effectively be reduced due to evaporation during the upward transportation process, which further alters the chemical compositions and the particle sizes. The transportation of the aerosols to a relatively cleaner environment prevails over the formation of secondary chemicals at a high altitude (D. T. Liu et al., 2020). Therefore, a comprehensive investigation of the detailed characteristics of aerosol formation and mixing states is required to understand their environmental effects at low and high altitudes.

As a typical high-altitude region, the Tibetan Plateau (TP) has the highest and largest mountain area in the world, which is the most sensitive and obvious indicator of climate change on the entire Asian continent (Liu et al., 2017; Chen and Bordoni, 2014; Immerzeel et al., 2010). Numerous studies have shown that the melting and retreat of glaciers in the TP region have been accelerating in recent decades, largely attributed to anthropogenic emissions, such as greenhouse gases and aerosols (Luo et al., 2020; Hua et al., 2019). Atmospheric aerosols can also act as cloud condensation nuclei to impact the local hydrological cycles and monsoon patterns by changing the microphysical properties and life span of clouds (Qian et al., 2011; Gettelman et al., 2013; Kumar et al., 2017). The southern part of the TP is always affected by the transport of more polluted air from South Asia along the mountain valleys, especially during the pre-monsoon (i.e. March–May) with the south-west-prevailing wind (Chan et al., 2017; Zhao et al., 2017; Han et al., 2020). Most studies have focused on the optical characteristics within the TP; however, only little research has been conducted on aerosol components.

Present research on aerosol components over the TP mostly focuses on exploring the influence of light-absorbing carbon aerosols and dust particles on climate change by optical or offline sampling methods (e.g. Q. Y. Wang et al., 2019; Liu et al., 2021). There is a lack of studies on the chemical composition, mixing states, and formation mechanism of aerosols in the south-eastern margin and even the entire TP, especially using high-time-resolution measurements. Although time-integrated sampling with filter collection followed by laboratory analyses has been widely adopted for the chemical characterization of aerosols (C. L. Li et al., 2022; Shen et al., 2015; Zhang et al., 2013), the drawbacks of the traditional approach need to be given attention, including the low time resolution, high detection limit, and time- and labour-intensive procedures. Therefore, more advanced aerosol measurement equipment with high time resolution has been developed; for example the aerosol chemical speciation monitor (ACSM) and aerosol mass spectrometer (AMS) (Ng et al., 2011; Canagaratna et al., 2007) mainly obtained the online data of non-refractory submicron aerosol (including the mass concentration of sulfate, nitrate, ammonium, chloride, and organic matter and the mass spectrum of organic matter). This is beneficial for recognizing the dynamic processes of source emissions of organic matter in the atmosphere (Du et al., 2015; X. Zhang et al., 2019). Meanwhile, aerosol time-of-flight mass spectrometry (ATOFMS) (Dall'Osto et al., 2014) and single-particle aerosol mass spectrometry (SPAMS) (Zhang et al., 2020) are popular for characterizing individual atmospheric particles. These devices can determine the chemical composition and size distribution of the particles in detail and further analyse the dynamic processes of chemical ageing, mixing state, and transport of the particles. (Liang et al., 2022; L. Li et al., 2022a; G. H. Zhang et al., 2019). To the best of our knowledge, the advanced measurement device has not yet been applied for the studies conducted in the TP, leading to a lack of in-depth research on PM_2.5 pollution in the TP, especially in the south-eastern margin, which hinders our understanding of the distribution characteristics and formation mechanism of aerosol components in high-altitude regions.

The south-eastern margin of the TP is an important transitional zone between the high-altitude TP and the low-altitude Yungui Plateau (Q. Y. Wang et al., 2019; Zhao et al., 2017) and is an ideal place for investigating the impacts of pollutant transport and formation in the high-altitude zone. In this study, continuous field observations of individual particles (SPAMS) were made in the south-eastern margin of the TP during the pre-monsoon period to (i) investigate the changes in chemical characteristics between transport and local fine particles during the pre-monsoon, (ii) determine the size distributions and mixing states of different particle types, and (iii) assess the contributions of photooxidation and aqueous reaction to the formation of the secondary species. These results can expand our understanding of the chemical components, size distribution, mixing state, and ageing pathways of aerosols in the high-altitude areas over the TP and surrounding areas.

2 Methodology

2.1 Sampling site

Intensive 1-month field observations were made on the rooftop (∼10 m above ground level) of the Lijiang Observatory, Chinese Academy of Sciences (3260 m above sea level; 26^∘41^′24^′′ N, 100^∘10^′48^′′ E), Gaomeigu County, Yunnan Province, China, during the pre-monsoon period (from 14 April to 13 May 2018). The nearest residential area is the village of Gaomeigu (3–5 km away), with a small population of 113 residents in 27 households. Villagers earn a living by farming (e.g. potato and autumn rape), and biomass is the major domestic fuel (Li et al., 2016). The sampling site is surrounded by rural and mountainous areas and has no obvious industry or traffic emissions. During the total observation period, the average temperature (T) and relative humidity (RH) were 8.4±3.1 ^∘C and 69±21 %, respectively. The wind speed (WS) was 2.2±1.2 m s⁻¹, with the prevailing wind in the north and north-east (Fig. S1 in the Supplement).

2.2 Online instrument

A detailed operational principle and the calibrations of the single-particle aerosol mass spectrometer (SPAMS; Hexin Analytical Instrument Co., Ltd., Guangzhou, China) have been described elsewhere (Li et al., 2011). Briefly, individual particles are drawn into the SPAMS through a critical orifice. The particles are focused and accelerated, then aerodynamically sized by two continuous diode Nd:YAG laser beams (532 nm), and subsequently desorbed and ionized by a pulsed laser (266 nm) triggered exactly based on the velocity of the specific particle. The generation of positive and negative molecular fragments is recorded with the corresponding size of individual particles. In summary, a velocity, a detection time, and an ion mass spectrum are recorded for each ionized particle, while there is no mass spectrum for non-ionized particles. The velocity could be converted to d_va based on a calibration using polystyrene latex (PSL) spheres (Thermo Scientific Corp., Palo Alto, USA) with predefined sizes. The average ambient pressure is 690 hPa (in a range of 685–694 hPa) during the measurements and calibration. A hollow silicone dryer was installed in front of the inlet. This reduces the uncertainty in particle collection efficiency due to the changes in humidity in sampled airs. Particles measured by the SPAMS mostly have a vacuum aerodynamic diameter (d_va) of 0.2–2.0 µm. This SPAMS-specific size distribution is semi-quantitatively evaluated by the relative concentration and contribution of each particle type, mainly due to its large dependence on the particle detection efficiency (Allen et al., 2000; Yang et al., 2017). The characteristics of the SPAMS-specific size distribution are statistical results, while the comparison of the relative distribution and number fraction of different particle types in each size bin is significant.

Meteorological parameters, including the temperature (^∘C), RH (%), WS (m s⁻¹), and wind direction (WD), were continuously measured using an automatic weather station (Model MAWS201, Vaisala HydroMet, Helsinki, Finland) at a 5 min resolution, and the planetary boundary layer (PBL) was acquired from the website https://doi.org/10.5065/D6M043C6 (National Centers for Environmental Prediction et al., 2000) at a 1 h resolution. Gaseous concentrations (ppbv) were obtained using a multiple gas analyser (Thermo Scientific Corp.), including ozone (O₃; Model 49i) and nitrogen oxides (NO_x; Model 42i) at a 5 min resolution. The SPAMS and gas analysers were co-located in the same position, and the weather station was uncovered outside ∼5 m from the sampling house. Time series of SPAMS particles, gaseous concentrations (NO, NO_x, O₃, and CO), and meteorological parameters (PBL, temperature, RH, WD, and WS) are shown in Fig. S2.

2.3 Individual particle classification

During the observation period, a total of 461 876 ambient particles with a size (d_va) of 0.2–2.0 µm were collected, including 55 583 in Episode 1 (E1; from 18 April at 08:00 local time (LT) to 19 April at 08:00 LT) and 62 110 in Episode 2 (E2; from 26 April at 17:00 LT to 28 April at 02:00 LT). The analysed particles are classified into 1557 groups using an adaptive resonance theory neural network (ART-2a) with a vigilance factor of 0.8, a learning rate of 0.05, and 20 iterations (Song and Hopke, 1999). Finally, eight major particle cluster types (i.e. potassium-rich (K-rich), biomass burning (BB), organic carbon (OC), ammonium, aged elemental carbon (aged EC), dust, sodium (Na)–potassium (K)–sulfate (S)–nitrate (N)-containing (NaK-SN), and iron (Fe)–lead (Pb)-containing (metal)) with distinct chemical patterns were manually combined, representing ∼99.7 % of the population of the detected particles. The remaining particles are grouped as “other”. The characteristics of the positive and negative mass spectra of each particle type are shown in Fig. S3. A detailed description of classification criteria for individual particles and the characteristic ion fragments for each particle type can be found in Sect. S1. The criteria used for searching some of the secondary species in the SPAMS datasets are summarized in Table S2 in the Supplement.

2.4 Trajectory-related analysis

To determine the influence of regional transport on different particles at the south-eastern margin of the TP, the trajectory cluster analysis was carried out using the 72 h backward air mass trajectories at arrival heights of 500 m above ground level. The trajectories were calculated with the Hybrid Single-Particle Lagrangian Integrated Trajectory model (Draxler and Hess, 1998), and the meteorological data were obtained from the Global Data Assimilation System (GDAS; ftp://arlftp.arlhq.noaa.gov/pub/archives/gdas1, last access: 6 April 2022). The cluster analysis employs a Euclidean-oriented distance definition to differentiate and cluster the major spatial features of the inputting trajectories. Details of the trajectory-clustering method can be found in Sirois and Bottenheim (1995). To investigate the effects of transport on the chemical characteristics of the individual particles, trajectories with particle number concentrations higher than the 75th percentile are considered to be pollution (Liu et al., 2021).

3 Results and discussion

3.1 Characteristics of particle composition

Table 1 summarizes the number concentrations, relative percentages, and characteristic ions of each particle type. The most dominant particle type in Gaomeigu during the pre-monsoon is K-rich, accounting for an average of 30.9 % of the total resolved particles, followed by BB (18.7 %), OC (12.8 %), ammonium (11.9 %), aged EC (10.9 %), and dust (10.7 %). Their characteristics of mass spectrum and possible sources are described in Sect. S1 in detail. Similar to the results of some studies in urban areas, the K-rich or carbon-containing types are the dominant particle types (15 %–50 %) (Xu et al., 2018; H. L. Wang et al., 2019; L. Li et al., 2022). Based on a combination of previous studies and the characteristics of the mass spectrum (Fig. S3a) in this study, the K-rich particles are contributed by biomass burning and traffic emissions, as evidenced by these extensive works usually identifying abundant ³⁹K⁺ signals for biomass burning (Pratt et al., 2011; Chen et al., 2017), while the presence of phosphate ($m/z$ ⁷⁹PO${}_{\mathrm{3}}^{-}$) indicates vehicle exhaust (Yang et al., 2017). The results of the correlation between seven variables (Fig. S4) show that the K-rich type is strongly correlated with the ammonium (r=0.84) and aged-EC (r=0.90) types and is well correlated with the OC (r=0.70) and BB (r=0.68) types, further demonstrating that the K-rich particle type is from traffic emissions and biomass burning and is affected by secondary formation during atmospheric ageing in the south-eastern TP. It is worth noting that little research has captured the high proportion of ammonium particles as shown in this study (Shen et al., 2017; Xu et al., 2018), which is ascribed to the conversion of the ammonia (NH₃) precursor emitted from large-scale agricultural activities and mountain forests (Engling et al., 2011; Li et al., 2013). It is necessary to point out that 60 % of ammonium particles contain signals of amine fragment ($m/z$ 58, ${\mathrm{C}}_{\mathrm{2}}{\mathrm{H}}_{\mathrm{5}}\mathrm{NH}={\mathrm{CH}}_{\mathrm{2}}^{+}$), implying their similar formation pathway (Zhang et al., 2012). Moreover, the amine-containing particles represented 12.5 % of the total ambient particles, which is significantly higher than in some urban areas at low altitudes (around 2 %) (Cahill et al., 2012; Zhang et al., 2015; Li et al., 2017) but is comparable to observed sites with high RH or during fog and cloud events at a high altitude (>9 %) (Roth et al., 2016; Lin et al., 2019). This suggests that the formation of amines under high-RH and high-fog conditions might exist in the Gaomeigu area (with an altitude of 3260 m); for example, the high relative fraction of amine-containing particles corresponds to a high RH (Fig. S5), and the existence of amine sources governs the ammonium formation (Bi et al., 2016; Rehbein et al., 2011). The relatively larger fraction of dust particles is related to the short occurrences of dust events in spring (Fig. S6), leading to a wide contribution ranging between 10 % and 70 % in the period of 19:00 LT on 16 April to 10:00 LT on 17 April.

Table 1 The number concentrations, average percentages, and characteristic ions of nine types of particles during the entire campaign and the average percentages of the six major particle types during two episodes.

NA: not available.

Figure 1 shows the diurnal variations in each particle type. The K-rich, BB, and OC particles decrease after midnight until 06:00 LT, possibly explained by the curtailment of local traffic and biomass burning activities at nighttime even though both the PBL height and WS decrease (Fig. S7). Then, their concentrations rapidly increase in the morning (around 07:00 LT) due to more pollutants from biomass burning and traffic emissions in the upwind region. The increases in PBL height and WS also lead to the transport of air pollutants from the surrounding regions to the sampling site (Liu et al., 2021). At 11:00 LT, the particle concentrations sharply decrease until 16:00–17:00 LT, caused by the pollutant dispersion with continuing increases in the PBL height and WS. Increasing trends are observed after 17:00 LT due to the pollutant accumulation with the reduction in PBL height and WS. In contrast, the ammonium, aged-EC, and dust particles show a unimodal pattern of the daily diurnal variation (Fig. 1d–f). From 00:00 to 06:00 LT, minor fluctuations in particle concentrations of ammonium, aged-EC, and dust particles are observed. After that, their levels continuously elevate until ∼ 11:00 LT due to the regional transport, traffic emissions, and fugitive dust (refer to Sect. S2). While the PBL height and WS increase continuously, the ammonium, aged-EC, and dust types decline from 12:00 to 17:00 LT. The subsequent increases in these three types after 17:00 LT are attributed to the reduction in PBL height as a result of the accumulation of pollutants in the near-surface atmosphere.

Figure 1 Diurnal box-and-whisker plots of the number concentration of the main particle types (at an hourly resolution): (a) potassium (K)-rich, (b) biomass burning (BB), (c) organic carbon (OC), (d) ammonium, (e) aged elemental carbon (EC), and (f) dust. The lower, middle, and upper lines of the boxes denote the 25th, 50th, and 75th percentiles. The lower and upper whiskers represent the 10th and 90th percentiles, respectively. Average values are shown by white dots.

Based on the transport pathways, four air mass clusters are identified to investigate the effect of regional transport on the major particle types (i.e. K-rich, BB, OC, ammonium, aged EC, and dust) (Fig. 2). Clusters 1, 3, and 4 originated from north-eastern Myanmar, accounting for 59.8 %, 33.2 %, and 4.6 % of the total trajectories, respectively. Cluster 1 had an average percentage of 32.7 %, 18.5 %, 12.0 %, 12.5 %, 11.1 %, and 8.9 %, respectively, for the K-rich, BB, OC, ammonium, aged-EC, and dust particles (Table S1). Clusters 3 and 4 have comparable contributions of OC (15.5 % and 12.5 %, respectively) and increased BB (19.3 % and 26.8 %, respectively) as well as decreased K-rich (26.8 % and 25.2 %, respectively), ammonium (10.4 % and 7.7 %, respectively), and aged-EC (7.7 % and 6.3 %, respectively) particles compared to those of Cluster 1, but with a high contribution of dust (16.6 %), which suggests that Clusters 3 and 4 are significantly correlated with dust and biomass burning pollution. However, Cluster 1 is more influenced by compound pollution, mainly including secondary formation, biomass burning, and traffic emissions. The diurnal variations in the BB and OC fractions are similar and rapidly elevate at 07:00 LT (Fig. S8) due to the increased contribution of biomass burning and traffic emissions from Cluster 1 and ammonium and aged-EC particles (peak at 07:00 LT) caused by the effect of Clusters 1 and 3 together. A stable diurnal variation in the K-rich fraction is mainly due to its large proportion and diverse sources. The similar diurnal trends of Clusters 3 and 4 are both associated with dust contributions, which decrease at 04:00 LT and increase at noon. The increased nighttime particles could be attributed to the pollutant accumulation with decreased PBL height. Cluster 2 originates from north-eastern India and passes over Bangladesh. This cluster accounts for only 2.4 % of the total trajectories, of which ∼30.8 % and ∼35.9 % are mainly associated with the K-rich and BB particles, respectively. Although Clusters 2 and 4 are composed of a small fraction of total trajectories (2.4 % and 4.6  %, respectively), BB and dust particles are identified as the major pollutants, suggesting significant influences from India and north-eastern Myanmar during the campaign.

Figure 2 Maps of the mean HYSPLIT back-trajectory clusters (72 h) at a height of 500 m during the whole field observation. The embedded pie chart represents the relative fraction of each particle type in the four clusters.

A more in-depth investigation of the characteristics of the main particle types in the south-eastern Tibet Plateau was conducted during two episodes when the number concentration of particles was high (E1: from 08:00 LT on 18 April to 08:00 LT on 19 April 2018; E2: 17:00 LT on 26 April to 02:00 LT on 28 April 2018) (Fig. S6). Even though the two episodes are caused by Cluster 1, the chemical components show significant differences (Table 1). During E1, the average fractions of the K-rich, BB, OC, ammonium, aged-EC, and dust particles are 29.0 %, 11.5 %, 8.1 %, 17.5 %, 10.0 %, and 20.3 %, respectively, different from 39.3 %, 14.2 %, 10.0 %, 13.5 %, 17.2 %, and 1.3 %, respectively, during E2. It can be seen that the major changed factor of the dust particles is 93.6 % lower during E2 than E1, and that of the opposite aged-EC particles is 72.0 % higher than E1. Meanwhile, K-rich, BB, and OC particles also increase by 35.5 %, 23.5 %, and 23.4 %, respectively, during E1 compared to E2. For the air mass clusters (Fig. S9), E1 and E2 exhibit minor differences, mostly originating from northern Myanmar and the Sino-Burmese border, but not identical regions. The dust particles that are much lower during E2 than E1 could be explained by higher WS (on average 2.7±1.0 m s⁻¹ versus 0.4±0.5 m s⁻¹) (Fig. S9) and PBL height (771±717 m versus 560±549 m) (Fig. S10). The dust particles are mainly formed by re-suspension in the local areas. In addition, quickly thrown-up dust belongs to the category of more coarse particles, which are out of the detection range of the SPAMS. However, because the larger dust particles deposited more easily under low-WS and stagnant-air conditions during E1, more suspended dust particles of small size fall within the detection range of the SPAMS. Moreover, the increased PBL height and WS could speed up the transportation of pollutants from multiple sources (e.g. traffic and biomass burning emissions) to the observation site, leading to elevation of the fraction of aged-EC, K-rich, BB, and OC particles during E2. The decreased ammonium fraction during E2 is potentially explained by the reductions in the secondary pollutant formation with declines in RH (from 73.9±23.9 % to 53.1±14.9 %), in comparison to those during E1.

3.2 Characteristics of the SPAMS-specific size distribution and mixing state

The SPAMS-specific size distributions of all particle types are shown in Fig. 3. According to the characteristics of the average mass spectra (refer to Sect. S1 and Fig. S3), the K-rich, BB, OC, and aged-EC particles originated from similar sources of solid-fuel combustion or vehicle emissions. Their SPAMS-specific size is thus distributed within a small scale (∼440 nm) (Fig. S11a). However, the relative percentage of each particle type is distinct, with different size ranges, possibly due to the unique atmospheric processing. For example, as shown in Fig. 3a, the proportions of the K-rich and BB types increase along with an increase in particle size from 200 to 420 nm, and then they decrease. The OC and aged-EC types are mainly distributed in relatively small sizes, and their proportions gradually decrease when the size ranges become larger. The ammonium and dust types are mainly distributed in large sizes of ∼600 nm (Fig. S11a). The proportion of ammonium particles gradually increases with an increase in particle size and peaks at 740 nm; the relatively large SPAMS-specific size distribution is ascribed to the intense atmospheric ageing during regional transport (refer to Sect. S1). The proportion of dust particles gradually increases with a size >560 nm and peaks at 1.48 µm. This is consistent with the fact that dust is a coarse particle, generally associated with fugitive dust.

Figure 3 SPAMS-specific size distributions of the relative number fraction (%) of the total particles for nine groups during (a) the total sampling campaign and the two episodes (b) E1 and (c) E2.

Compared with the SPAMS-specific size distribution of the total particles, the peak values of the six main particle types show minor differences (<80 nm) during the two different episodes (Fig. 11b, c). However, the percentage of the six particle types is distributed in wider size ranges during E2 than during E1, possibly due to the more intense atmospheric ageing. Similarly, during the two episodes (Fig. 3b, c), the relatively high fractions of the K-rich and BB particles are more affected by the primary emissions when their peak values of the SPAMS-specific size distribution concentrate at <300 nm, and size distributions >300 nm are more related to the ageing process (L. Li et al., 2022a; Bi et al., 2011). Relatively greater fluctuation for the large-sized fractions (>1.1 µm) could be explained by the low particle concentration (a number less than 20). It should be pointed out that further application of this method would require a co-located particle-sizing instrument to scale the size-resolved particle detection efficiency. Both particle composition and size dependence are the predominant impacting factors in the particle detection efficiency of the SPAMS (Wenzel et al., 2003; Yang et al., 2017; Healy et al., 2013).

To investigate the mixing state of the secondary species in the six main particle types, the number fractions of six secondary markers (⁹⁷HSO${}_{\mathrm{4}}^{-}$, ¹⁹⁵H(HSO${}_{\mathrm{4}}{)}_{\mathrm{2}}^{-}$, ⁶²NO${}_{\mathrm{3}}^{-}$, ¹⁸NH${}_{\mathrm{4}}^{+}$, ⁵⁸C₂H₅NHCH${}_{\mathrm{2}}^{+}$, and ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$) are calculated (Fig. 4). The presence of amine ($m/z$ ⁵⁸C₂H₅NHCH${}_{\mathrm{2}}^{+}$) and sulfuric acid ($m/z$ ¹⁹⁵H(HSO${}_{\mathrm{4}}{)}_{\mathrm{2}}^{-}$) signals is possibly indicative of the water uptake (Chen et al., 2019) and acidic properties of the particles (Rehbein et al., 2011), respectively. The mixing states are obtained by the ratio of the number concentration of the selected ions in each particle type.

Figure 4 Number fractions of secondary markers associated with the six particle types (K-rich, BB, OC, ammonium, aged EC, dust) during the whole observation. Secondary species include sulfate (⁹⁷HSO${}_{\mathrm{4}}^{-}$), sulfuric acid (¹⁹⁵H(HSO${}_{\mathrm{4}}{)}_{\mathrm{2}}^{-}$), nitrate (⁶²NO${}_{\mathrm{3}}^{-}$), ammonium (¹⁸NH${}_{\mathrm{4}}^{+}$), amine (⁵⁸C₂H₅NHCH${}_{\mathrm{2}}^{+}$), and oxalate (⁸⁹HC₂O${}_{\mathrm{4}}^{-}$) ions.

The most abundant ⁹⁷HSO${}_{\mathrm{4}}^{-}$ and ¹⁸NH${}_{\mathrm{4}}^{+}$ fractions are seen in ammonium (99 % and 94 %, respectively) and aged-EC (92 % and 31 %, respectively) particles, whereas a very low fraction of ⁶²NO${}_{\mathrm{3}}^{-}$ is found (2 % and 7 %, respectively). These values indicate that ammonium sulfate is the predominant form rather than ammonium nitrate. (Zhang et al., 2013). The high contribution of ⁹⁷HSO${}_{\mathrm{4}}^{-}$ in EC-containing particles also suggests a significant influence of anthropogenically emitted sulfate precursors (e.g. SO₂) on the ageing of EC-containing particles at high altitudes (Peng et al., 2016; J. K. Zhang et al., 2017). Meanwhile, relatively high number fractions of ¹⁹⁵H(HSO${}_{\mathrm{4}}{)}_{\mathrm{2}}^{-}$ and ⁵⁸C₂H₅NHCH${}_{\mathrm{2}}^{+}$ are also observed in ammonium (63 % and 60 %) and aged-EC (4 % and 19 %) particles. These abundant mixtures potentially represent the high hygroscopicity of ammonium and aged-EC particles and their ability to neutralize the acidic particles of ammonium particles (Sorooshian et al., 2007). Then, a moderate fraction of ⁹⁷HSO${}_{\mathrm{4}}^{-}$ and ¹⁸NH${}_{\mathrm{4}}^{+}$ is seen in the K-rich (65 %, 7 %) and OC (56 %, 4 %) particles. In contrast, a higher ⁶²NO${}_{\mathrm{3}}^{-}$ fraction contributes to the K-rich (38 %) and OC (68 %) particles, mainly affected by vehicle emissions and biomass burning (refer to Sect. S1). A relatively low fraction of ⁹⁷HSO${}_{\mathrm{4}}^{-}$ contributes to BB (18 %) and OC (6 %) particles, and the moderate ⁶²NO${}_{\mathrm{3}}^{-}$ fraction mixes in BB (45 %) particles but only accounts for 3 % of dust particles. Combined with the results of the minor ¹⁸NH${}_{\mathrm{4}}^{+}$ fraction (<1 %) in BB and dust particles, this suggests a relatively low degree of ageing. In addition, oxalate (⁸⁹HC₂O${}_{\mathrm{4}}^{-}$), a representative component of secondary organic formation, is mainly mixed with BB (13 %) and K-rich (12 %) particles. This is because the substantial precursors of oxalic acid, including acetate (⁵⁹C₂H₃O${}_{\mathrm{2}}^{-}$), methylglyoxal (⁷¹C₃H₃O${}_{\mathrm{2}}^{-}$), and glyoxylate (⁷³C₂HO${}_{\mathrm{3}}^{-}$), are emitted from biomass burning, and then oxalate heterogeneously forms in BB-related particles (G. H. Zhang et al., 2019; Zauscher et al., 2013). A relatively low fraction (<5 %) of oxalate-containing particles in OC, ammonium, aged-EC, and dust particles is potentially limited by the contributions of precursor oxalic acid.

Compared to the mixing state of individual particles in urban or suburban areas that are located close to emission sources (Chen et al., 2016; Dall'Osto and Harrison, 2012; J. K. Zhang et al., 2017; L. Li et al., 2022a), the high fractions of sulfate and ammonium in high-altitude areas demonstrate a high degree of ageing of the individual particles, whereas the low fraction of nitrate with high volatility indicates its loss during transportation processing.

The number fractions of six markers in the four trajectories were used to further investigate the impacts of regional transport. As shown in Fig. 5a and c, the dominant mixing ion types in each particle (except for dust) are similar among the four clusters. For Cluster 1, the number fractions of ⁹⁷HSO${}_{\mathrm{4}}^{-}$ and ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$ have larger values in five particle types (except for the dust type) than those in other trajectories. Similar to Cluster 1, Clusters 3 and 4 are impacted by regional transport from north-eastern Myanmar, and the fractions of the six markers are also similar in the OC, ammonium, and aged-EC types. However, ⁹⁷HSO${}_{\mathrm{4}}^{-}$ in Clusters 3 and 4 is reduced in the K-rich, BB, and dust types, while ⁶²NO${}_{\mathrm{3}}^{-}$ is increased in the K-rich type and decreased in the dust type, compared with Cluster 1. As discussed in Sect. 3.1, these results demonstrate that the ageing degree of Clusters 3 and 4 might be lower than that of Cluster 1. For Cluster 2, the fraction of ⁹⁷HSO${}_{\mathrm{4}}^{-}$ is obviously decreased in the K-rich, BB, and aged-EC types but slightly increased in the dust type (Fig. 5f). However the fraction of ⁶²NO${}_{\mathrm{3}}^{-}$ is increased in the K-rich, OC, and dust particles in Clusters 2 compared with Clusters 1, 3, and 4. These variations in Cluster 2 are more likely due to the influences of biomass burning activities from surrounding the sampling site rather than regional transport. Furthermore, Cluster 2 is associated with regional transport from north-eastern India from the afternoon to nighttime (from 15:00 LT on 11 May to 07:00 LT on 12 May), which is favourable to the nitrate formation of N₂O₅ by heterogeneous hydrolysis (Wang et al., 2017; Ding et al., 2021). However, these cases are infrequent, as only 2 % of trajectories are associated with Cluster 2.

Figure 5 Number fractions of secondary markers associated with the six particle types (i.e. K-rich, BB, OC, ammonium, aged EC, and dust) in four clusters. Secondary species abbreviations as in Fig. 4.

During E1, ⁹⁷HSO${}_{\mathrm{4}}^{-}$ fractions greater than 50 % are mixed in the K-rich (81 %), OC (62 %), ammonium (100 %), and aged-EC (98 %) particles (Fig. S12) as well as the low-BB (37 %) and dust (4 %) particles. Unlike with E1, the number fraction of ⁹⁷HSO${}_{\mathrm{4}}^{-}$ in dust increases to 34 % during E2, potentially associated with the enhancement by secondary formation during regional transport. However, the mixing states of ¹⁹⁵H(HSO₄)${}_{\mathrm{2}}^{-}$, ⁶²NO${}_{\mathrm{3}}^{-}$, NH${}_{\mathrm{4}}^{+}$, and oxalate fractions are similar between the two episodes. The ⁵⁸C₂H₅NHCH${}_{\mathrm{2}}^{+}$ fractions are significantly higher in E2 than E1 for ammonium (67 % versus 31 %) and aged-EC particles (48 % versus 17 %), due to the relatively higher-hygroscopicity behaviour (i.e. RHs) (Sorooshian et al., 2007).

3.3 Formation process of the high-particle-number-concentration episodes

Photochemical oxidation and aqueous-phase reaction are the key formation pathways of secondary species (Link et al., 2017; Xue et al., 2014; Jiang et al., 2019). The oxidant O_x (O₃+NO₂) concentration and RH usually serve as indicators of the degree of photochemical oxidation (Wood et al., 2010) and aqueous-phase reaction (Ervens et al., 2011), respectively, though the current O_x and RH conditions obtained using the in situ measurement are not indicative of the past conditions experienced by the particle. Thus, the relative number fractions of ⁴³C₂H₃O⁺-, ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$-, ⁶²NO${}_{\mathrm{3}}^{-}$-, ⁹⁷HSO${}_{\mathrm{4}}^{-}$-, and ¹⁸NH${}_{\mathrm{4}}^{+}$-containing particles of the total detected particles were selected to provide a rough estimate of the secondary-formation mechanism under ambient TP conditions (Liang et al., 2022). The correlations of the number fraction of each secondary species with the O_x concentrations (O_x) during daytime (from 06:00 to 20:00 LT) and RH during nighttime (from 20:00 to 06:00 LT the next day) are used to reveal the formation pathways during the two episodes (L. Li et al., 2022).

As illustrated in Fig. 6, for E1, ⁴³C₂H₃O⁺, ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$, ⁹⁷HSO${}_{\mathrm{4}}^{-}$, and ¹⁸NH${}_{\mathrm{4}}^{+}$ show significant negative linear correlations with O_x (p<0.01), and the correlation strengths range from moderate to strong (r= −0.51 to −0.81). However, the ⁶²NO${}_{\mathrm{3}}^{-}$ fraction shows an upward trend with an insignificant correlation (r=0.33, p>0.05) with the increase in O_x concentration. For E2, ⁴³C₂H₃O⁺ shows weak correlation with O_x (r=0.37, p>0.05) but strong correlations with ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$, ⁹⁷HSO${}_{\mathrm{4}}^{-}$, and ¹⁸NH${}_{\mathrm{4}}^{+}$ (r= 0.81–0.92, p<0.01). It should be noted that ⁶²NO${}_{\mathrm{3}}^{-}$ has a strong negative correlation (r= −0.85, p<0.01) with O_x. In general, the opposite linear relationship between secondary aerosol and O_x during E1 and E2 might be influenced by (i) the relatively low secondary formations because of the small number of precursors emitted from anthropogenic activities around the sampling site (Li et al., 2016); (ii) the higher dilution rate of the particles formed in the atmosphere with the more rapid rise in PBL height during E1 than E2 (Fig. S13a); and (iii) the degrees of contribution of regional transport due to the low WS (0.5±0.6 m s⁻¹) during E1 and the high WS (3.1±1.0 m s⁻¹) during E2, respectively (Fig. S9). Therefore, for E1, the increases in the NO${}_{\mathrm{3}}^{-}$ fraction could be influenced by the local nitrate formation, while the declines in other secondary components should be ascribed to the reduced contribution of regional transport. For E2, the decrease in the NO${}_{\mathrm{3}}^{-}$ fraction could be caused by the relatively higher volatilization loss of nitrate than other components through the regional transport. Additionally, previous work proves that the formations of organic nitrate species (such as ²⁷CHN⁺, ³⁰NO⁺, ⁴³CHO₁N⁺, and CHO_xN⁺) through the NO+RO₂ pathway dominate 80 % of the total nitrate production in tropical forested regions during summertime (Alexander et al., 2009). Aruffo et al. (2022) also found that low NO_x (i.e. <6 ppbv), compared to 2.3±0.8 ppbv in this study, could even promote the particle-phase partitioning of the lower volatility of organonitrates. These results suggest that the secondary organic species have different formation capacities through photooxidation reactions, among which the rate of HSO${}_{\mathrm{4}}^{-}$ formation (slope =0.017) is the highest. With increased O_x concentration during E2, the concentration levels of secondary organic species of C₂H₃O⁺ (18 %–28 %) imperceptibly rise, while the oxalate fraction significantly increases by 7 %–20 %.

Figure 6 Correlations between the relative number fractions of the secondary species (a) ⁴³C₂H₃O⁺, (b) ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$, (c) ¹⁸NH${}_{\mathrm{4}}^{+}$, (d) ⁶²NO${}_{\mathrm{3}}^{-}$, and (e) ⁹⁷HSO${}_{\mathrm{4}}^{-}$ and O_x concentration during E1 (blue square) and E2 (red dot).

Considering that the oxalate is abundantly mixed in K-rich (14 %), BB (15 %), aged-EC (5 %), and dust (6 %) particles in Cluster 1 (Fig. 5) and considering the increased contributions of the K-rich (39.3 %), BB (14.2 %), and aged-EC (17.2 %) types during E2 (Table 1), the apparent formation of oxalate might be due to the enhancement of regional transport. Particularly, this shows that the nearby biomass burning and combustion activities produce more precursor species of oxalate (Sullivan et al., 2007; Kundu et al., 2010; G. Zhang et al., 2017).

Figure 7 illustrates that the number fractions of ⁴³C₂H₃O⁺, ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$, ⁹⁷HSO${}_{\mathrm{4}}^{-}$, and ¹⁸NH${}_{\mathrm{4}}^{+}$ have moderate to strong positive correlations with RH (r= 0.70–0.81, p<0.01 or 0.05) in the nighttime during the two episodes, except ⁴³C₂H₃O⁺ during E2 (p=0.48) and ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$ during E1 (p=0.12). Furthermore, the ⁶²NO${}_{\mathrm{3}}^{-}$ fraction has no obvious changes, with insignificant correlation with RH during E1 (p=0.43), and presents a moderate negative correlation with RH (r=0.69, p<0.01) during E2. As shown in Fig. 7e, the highest aqueous formation rate of HSO${}_{\mathrm{4}}^{-}$ is mainly due to the low volatility and highly hygroscopic properties of sulfate (G. H. Wang et al., 2022; S. P. Zhang et al., 2019; Sun et al., 2013). Compared with that during E2 (slope =0.014), the decreased formation rate of HSO${}_{\mathrm{4}}^{-}$ during E1 (slope =0.009) may be because of the decreases in aerosol acidity with higher RH (>80 %; Huang et al., 2019; Meng et al., 2014; Tian et al., 2021). And the increased contributions of regional transport due to the high WS (2.4±0.8 m s⁻¹) during E2 are compared with the low WS (0.08±0.08 m s⁻¹) during E1 (Fig. S9). The fair production rates of NH${}_{\mathrm{4}}^{+}$ during E1 (slope =0.005) and E2 (slope =0.006) demonstrate that an aqueous-phase reaction could effectively promote ammonium formation. Meanwhile, a slightly larger slope of NH${}_{\mathrm{4}}^{+}$ during E2 could also be affected by the increased contributions of regional transport. Compared with those during E1, the inverse generation rates of two secondary organic species (i.e. C₂H₃O⁺ and HC₂O${}_{\mathrm{4}}^{-}$) during E2 are possibly caused by the different formation pathways with a variety of RH levels or distinct regional transports. For example, C₂H₃O⁺ shows a strong correlation with RH (r=0.70, p<0.05) during E1 (slope =0.003) but has an insignificant correlation during E2. This could be explained by high RHs that could effectively promote secondary organic formation during E1. In addition, the HC₂O${}_{\mathrm{4}}^{-}$ fraction increases slightly (9.7 %–13.1 %) during E1, which is potentially ascribed to more abundant dust-type particles (20.3 %), which are composed of high calcium (Ca) content (Fig. S14) and favour the formation of metal oxalate complexes (i.e. Ca oxalate). At high RHs (93.4±7.6 %), if oxalate ions are dissolved in the aqueous phase with the presence of Ca ions, the Ca oxalate complexes can precipitate because of their low hygroscopicity and insoluble nature (Furukawa and Takahashi, 2011). This could offset the oxalate formation in the aqueous-phase reaction. However, significant linear increases (slope =0.003) with RH (r=0.81, p<0.01) during E2 demonstrate that the aqueous-phase reaction effectively promotes the oxalate formation (Cheng et al., 2017; Meng et al., 2020). No significant correlation between ⁶²NO${}_{\mathrm{3}}^{-}$ and RH is found during E1, potentially attributed to the decreases in NO₂ concentration (3.7±0.4 ppbv) in the local atmosphere. Meanwhile, high RHs could promote organonitrate formation (Fang et al., 2021; Fry et al., 2014). The linearity between ⁶²NO${}_{\mathrm{3}}^{-}$ and RH (r=0.69, p<0.01) significantly decreases during E2, mostly due to the losses of the volatile compound through regional transport (Fig. S15).

Figure 7 Correlations between the relative number fractions of the secondary species (a) ⁴³C₂H₃O⁺, (b) ⁸⁹HC₂O${}_{\mathrm{4}}^{-}$, (c) ¹⁸NH${}_{\mathrm{4}}^{+}$, (d) ⁶²NO${}_{\mathrm{3}}^{-}$, and (e) ⁹⁷HSO${}_{\mathrm{4}}^{-}$ and relative humidity (RH) during E1 (cyan dot) and E2 (orange square).

4 Conclusions

This study presents the chemical composition, size distribution, mixing state, and secondary formation of individual particles in the south-eastern margin of the TP, China, during the pre-monsoon season using a high-resolution SPAMS. The finding shows that the K-rich (30.9 %) and BB types (18.7 %) are the two dominant aerosol particles in the remote area, followed by the OC (12.8 %), ammonium (11.9 %), aged-EC (10.9 %), and dust (10.7 %) types; the NaK-SN, metal, and other particle types contributed 0.3 %–2.8 % of the total ambient particles. By interpreting the mass spectra and diurnal trends, the major particle types are mainly from traffic emissions, biomass burning, secondary formation, and fly ash, while the dynamics of the PBL height could also affect their contributions. The observed change in the number fraction of the particle types was mainly influenced by air mass (97.61 % of the total trajectories) from north-eastern Myanmar and significantly contributed to the K-rich and BB types. The particle types show distinct size distributions. The two critical particle types, the K-rich and BB types, appear in a unimodal pattern; the fractions of OC and aged-EC particles gradually decrease with the increase in particle size, but the ammonium and dust types show the opposite trend. Sulfate is the major secondary species and is highly mixed with the K-rich, ammonium, and aged-EC types. Nitrate has a relatively low mixing ratio due to its higher volatility than sulfate during regional transportation, while the relatively high fraction of nitrate in the BB and OC types is mainly affected by the sources of vehicle exhaust and biomass burning as well as ageing degree. During the entire study campaign, two episodes with high number concentrations of particles occur but with significant differences in each particle fraction due to the different meteorological conditions (RH, WS, etc.). Meanwhile, the different meteorological conditions also lead to an inverse linear correlation between the indicators of secondary formation, including C₂H₃O⁺, HC₂O${}_{\mathrm{4}}^{-}$, NH${}_{\mathrm{4}}^{+}$, NO${}_{\mathrm{3}}^{-}$, and HSO${}_{\mathrm{4}}^{-}$ as well as O_x (O₃+NO₂) during the periods of Episodes 1 and 2; however, they present a positive linear correlation with relative humidity (RH), except for NO${}_{\mathrm{3}}^{-}$, which shows a negative linear correlation with RH due to the low precursor concentration and potential organonitrate formation. These results demonstrate that the capacity of atmospheric ageing of photooxidation and aqueous reaction has complex influencing factors. Although the detailed formation pathways and their percentage contributions to secondary species are not quantitatively estimated in this study, our results have important implications for the various possibilities affecting the characteristics of chemical components, size distribution, mixing states, and formation mechanism of aerosols in the south-eastern TP. More in-depth investigations concerning the evolution mechanisms of secondary aerosols are encouraged since the TP is a significant regulator of global climate change.