New particle formation induced by anthropogenic–biogenic interactions on the southeastern Tibetan Plateau

New particle formation (NPF) plays a crucial role in the atmospheric aerosol population and has significant implications on climate dynamics, particularly in climate-sensitive zones such as the Tibetan Plateau (TP). However, our understanding of NPF on the TP is still limited due to a lack of comprehensive measurements and verified model simulations. To fill this knowledge gap, we conducted an integrated study combining comprehensive field measurements and chemical transport modeling to investigate NPF events on the southeastern TP during the pre-monsoon season. NPF was observed to occur frequently on clear-sky days on the southeastern TP, contributing significantly to the cloud condensation nuclei (CCN) budget in this region. The observational evidence suggests that highly oxygenated organic molecules (HOMs) from monoterpene oxidation participate in the nucleation on the southeastern TP. After updating the monoterpene oxidation chemistry and nucleation schemes in the meteorology–chemistry model, the model well reproduces observed NPF and reveals an extensive occurrence of NPF across the southeastern TP. The dominant nucleation mechanism is the synergistic nucleation of sulfuric acid, ammonia, and HOMs, driven by the transport of anthropogenic precursors from South Asia and the presence of abundant biogenic gases. By investigating the vertical distribution of NPF, we find a significant influence of vertical transport on the southeastern TP. More specifically, strong nucleation near the surface leads to an intense formation of small particles, which are subsequently transported upward. These particles experience enhanced growth to larger sizes in the upper planetary boundary layer (PBL) due to favorable conditions such as lower temperatures and a reduced condensation sink. As the PBL evolves, the particles in larger sizes are brought back to the ground, resulting in a pronounced increase in near-surface particle concentrations. This study highlights the important roles of anthropogenic–biogenic interactions and meteorological dynamics in NPF on the southeastern TP.

1 Introduction

The Tibetan Plateau (TP) features the highest and most extensive highland in the world. Referred to as the “Third Pole”, it exerts substantial impacts on global atmospheric circulation and the Asian monsoon climate (Qiu, 2008; Duan and Wu, 2005; Yanai et al., 1992; Liu et al., 2007, 2023). The thermal forcing from the TP during spring contributes to the seasonal transition of East Asian circulation, playing a crucial role in determining the onset site and timing of the Asian summer monsoon (Liu et al., 2007; Yanai et al., 1992; Hsu and Liu, 2003; Sato and Kimura, 2007; Liang et al., 2005; Wu and Zhang, 1998). Additionally, the TP, serving as the “Asian water tower”, is widely recognized as one of the most ecologically important areas (Lau et al., 2006; Xu et al., 2009; Lau et al., 2010; Immerzeel et al., 2010). The substantial climate warming trend has been occurring over the TP since the last half century, altering the atmospheric circulation over half the planet and the water supply for billions of people (Niu et al., 2004; Yao et al., 2019; Chen et al., 2015; Lau et al., 2010; Immerzeel et al., 2010; Qiu, 2008). Of all the possible causes, clouds exert an important role in rapid warming on the TP, as they influence radiative balance and latent heating or cooling resulting from water-phase changes (Ramanathan et al., 1989; Duan and Wu, 2006; Kang et al., 2010).

New particle formation (NPF) through gas–particle conversion is a vital contribution to cloud condensation nuclei (CCN) and thus significantly affects the cloud properties, especially in pristine areas (Merikanto et al., 2009; Wang and Penner, 2009; Yu and Luo, 2009; Kerminen et al., 2018). In most environments, sulfuric acid (H₂SO₄) is believed to play a crucial role in NPF due to its low volatility (Kulmala et al., 2013). However, it has been observed that H₂SO₄ alone cannot fully explain the observed formation rates in the lower atmosphere (Kirkby et al., 2011). The efficient clustering of H₂SO₄ with bases has been reported (Almeida et al., 2013). Nucleation processes involving H₂SO₄ and stabilizing species such as ammonia and amines are considered the dominant mechanism for NPF in urban and rural environments (Yao et al., 2018; Lai et al., 2022b; Yan et al., 2021). Additionally, highly oxygenated organic molecules (HOMs), formed through the oxidation of biogenic volatile organic compounds (BVOCs), are recognized as important drivers for NPF, especially under pristine or preindustrial conditions (Gordon et al., 2016, 2017). The nucleation of pure biogenic organic species in the absence of H₂SO₄ was observed in the Cosmics Leaving Outdoors Droplets (CLOUD) chamber (Kirkby et al., 2016). This mechanism is further supported by the measurements at the high-altitude sites (Bianchi et al., 2021, 2016). Notably, in forested areas and chamber experiments with the presence of both H₂SO₄ and BVOCs, H₂SO₄ clusters containing both base molecules and oxygenated organic molecules have been observed, suggesting the importance of multicomponent acid–base–organic nucleation (H₂SO₄–NH₃–organics–H₂O) (Lehtipalo et al., 2018). Moreover, higher concentrations of ions at high altitudes could assist the above mechanisms (Kirkby et al., 2016; Yu, 2010; Frege et al., 2017). In general, various nucleation mechanisms, which are highly sensitive to the concentrations of precursor vapors and air ions, make the NPF complicated in the ambient.

Given the TP's pristine nature and limited anthropogenic activities, NPF is the main source of atmospheric aerosols and CCN (Carslaw et al., 2013; Gordon et al., 2017). Yet, our knowledge of NPF over the TP, especially the mechanisms of nucleation, is hampered due to the lack of comprehensive observations (Ma et al., 2008). A few ambient observations of NPF have been conducted at high-altitude sites over the TP. Neitola et al. (2011) showed that NPF takes place frequently in the pre-monsoon season at Mukteshwar (2180 m a.s.l.) in the western Himalayas. This is attributed to the lifting of the boundary layer height, allowing the transport of precursor gases, such as SO₂, from polluted lower-altitude regions to the station. Similarly, Moorthy et al. (2011) found a high probability of NPF from precursor gases, potentially advected from inhabited regions, when solar insolation was abundant at the Hanle site (4520 m a.s.l.) in the Transhimalaya. A previous study conducted at the Himalayan Nepal Climate Observatory, Pyramid (NCO-P; 5079 m a.s.l.), located in the eastern Himalaya, revealed that NPF events occurred frequently when more polluted air from valleys reached the site (Venzac et al., 2008). Recent comprehensive measurements at the NCO-P found a high frequency of NPF during cleaner conditions in the absence of H₂SO₄, suggesting that aerosol production in this region mainly occurs through organic precursors of biogenic origin (Bianchi et al., 2021).

Serving as a moisture transport bridge between the South Asian monsoon and the East Asian monsoon, the southeastern TP (average altitude is about 4200 m a.s.l.) is a climate-sensitive zone for modulating atmospheric circulation (Li et al., 2016). Notably, the southeastern TP is situated at the convergence of significant natural and anthropogenic aerosol sources, including extensive alpine forests on a regional scale and densely populated areas to the south. Substantial BVOCs from alpine forests could undergo fast oxidation over the TP (Lin et al., 2008), favorable for the generation of HOMs. In addition to the natural process, anthropogenic emissions from the surrounding South Asia regions can also affect the TP through transport, contributing to the NPF-relevant precursors in the atmosphere (Zhao et al., 2013; Xia et al., 2011). The interaction between biogenic and anthropogenic sources adds further complication as well as interest to the NPF on the southeastern TP. Model simulations constrained by available in situ measurements offer a promising approach to gaining a deeper understanding of the NPF process. To shed more light on NPF on the southeastern TP, we conducted an investigation focusing on NPF events during the pre-monsoon period, which is of paramount importance from the point of view of clouds and the monsoon circulation. By integrating comprehensive in situ measurements in Lulang and updated regional chemical model simulations incorporating process diagnoses, we aim to provide a picture of NPF from a three-dimensional perspective over the southeastern TP.

2 Materials and method

2.1 Measurement site and in situ instrumentation

In this study, intensive in situ measurements were conducted from 4 April to 24 May 2021 at the Southeast Tibet Observation and Research Station (29°46^′ N, 94°44^′ E; 3200 m a.s.l.), a remote, high-altitude site in Lulang of the southeastern TP. During the pre-monsoon season, this region is under the control of the southern branch of the westerlies with a prominent southerly wind (Fig. 1a). In general, the measurement site is characterized by extensive emissions of BVOCs, especially monoterpenes (Luo et al., 2002; Liang et al., 2011; Wang et al., 2007) (Fig. 1a). Despite a single nearby cottage that emits pollutants by burning wood at almost regular morning hours, anthropogenic emissions in Lulang were relatively low on a regional scale (Fig. 1b).

Figure 1 Spatial distribution of (a) biogenic monoterpene emission rates together with 700 hPa wind fields in March–April–May and (b) anthropogenic emission of SO₂ together with topographic field (purple isolines with the unit of kilometers). Note that the biogenic monoterpene emission rates were calculated online by using the Model of Emissions of Gases and Aerosols from Nature (MEGAN) that is embedded in WRF-Chem (Guenther et al., 2006). The anthropogenic SO₂ emission rates were obtained from the mosaic Asian anthropogenic emission inventory (MIX), with specific emissions for China sourced from the Multi-resolution Emission Inventory for China (MEIC) at 0.25° × 0.25° horizontal resolution (Li et al., 2017b). The black star marks the location of the Lulang site. The solid black line is the boundary of the Tibetan Plateau. The dashed pink lines in Fig. 1a define the domain of Fig. 1b.

The particle number size distributions in the size range from 1 nm to 20 µm were collectively measured using five instruments, including a particle size magnifier (PSM; Airmodus Inc.), a neutral cluster and air ion spectrometer (NAIS; Ariel Inc.), two scanning mobility particle sizers (nano-SMPS and long-SMPS; TSI Inc.), and an aerodynamic particle sizer (APS; TSI Inc.). The PSM measures the size distribution of particles in the size range from 1 to 3 nm (Vanhanen et al., 2011). The NAIS detects the air ion number size distribution from 0.8 to 40 nm and particle number size distribution from 2 to 40 nm (Mirme and Mirme, 2013). The nano-SMPS and long-SMPS measure particle size distributions over the size range of 4–70 and 12–540 nm, respectively. The APS measures the particle size distribution from 500 nm to 20 µm in the aerodynamic diameter. The monoterpene concentration was measured by a proton transfer reaction time-of-flight mass spectrometer (PTR-TOF-MS; IONICON Analytik Inc.). A nitrate chemical ionization instrument with a atmospheric pressure interface time-of-flight mass spectrometer (CI-APi-TOF; Aerodyne Research Inc.) was used to detect the H₂SO₄ and HOMs (Jokinen et al., 2012). Moreover, on a typical NPF day (day with a typical “banana” shape on the particle number size distribution temporal surface plot), such as 29 April 2021, the CI-APi-TOF was switched to the APi-TOF negative mode to detect the chemical composition of atmospheric cluster ions (Junninen et al., 2010). The PM_2.5 concentration was observed using the online analyzer (SHARP5030; Thermo Fisher Scientific Inc.). The multi-angle absorption photometer (MAAP; Thermo-Scientific Inc.) was utilized to measure the black carbon (BC) mass concentration. The SO₂ and O₃ concentrations were observed using the API T100 and T400 (Teledyne API Inc.), respectively. The meteorological parameters, such as air temperature, wind speed, and directions, were measured by meteorological sensors (WXT530; Vaisala Inc.). The details regarding instrumentation can be found in the Supplement (Sect. S1).

2.2 WRF-Chem model configuration

To quantitatively understand the NPF process on the southeastern TP, the Weather Research and Forecasting model coupled with Chemistry (WRF-Chem) model version 3.9 is employed in this work. WRF-Chem is a widely used chemical transport model, which considers a variety of online-coupled meteorological processes and chemical processes, such as the emission and deposition of pollutants, advection and diffusion, gaseous and aqueous chemical transformation, and aerosol chemistry and dynamics (Grell et al., 2005). In this study, the simulation domain covers the southeastern TP, with a grid resolution of 27 km and 30 vertical layers from the ground level to the top pressure of 100 hPa. Although the complex terrain of the TP poses challenges for the simulations of fine-scale features (Sato et al., 2008; Zhang et al., 2020; Li et al., 2022), the choice of a 27 km and 30 layers resolution represent a compromise between computational efficiency and the need to capture key regional-scale atmospheric processes supported by previous studies (Li et al., 2015; Shi et al., 2008; Gao et al., 2018; Yang et al., 2018; Xu et al., 2018a; Rai et al., 2022; Ding et al., 2016; Huang et al., 2020). The simulation was conducted for 24–30 April, while the results of 28–30 April were analyzed to allow for the spinup for the chemical initial condition. The initial and boundary conditions of meteorological fields were constrained by the 6-hourly 1° × 1° National Centers for Environmental Prediction (NCEP) global final analysis (FNL) data. Moreover, NCEP Automated Data Processing (ADP) operation global surface observations and global upper-air observational weather data were assimilated to improve the simulation of meteorological fields. The chemical initial and boundary conditions are provided by the Whole Atmosphere Community Climate Model (WACCM) (Marsh et al., 2013). The WRF-Chem model simulations consider both natural and anthropogenic emissions. The anthropogenic emissions, including those from power plants, residential combustion, industrial processes, on-road mobile sources, and agricultural activities, were obtained from the mosaic Asian anthropogenic emission inventory (MIX) database at a monthly temporal resolution for the year 2010 (Li et al., 2017b), with specific emissions for China sourced from the Multi-resolution Emission Inventory for China (MEIC) at a monthly temporal resolution for the year 2017 (Zheng et al., 2018). Emission estimates from these inventories are subject to uncertainties due to the lack of complete knowledge of human activities and emissions from different sources (Li et al., 2017a, b; Zheng et al., 2018), which could cause biases in NPF simulations by misrepresenting the atmospheric oxidation capacity and condensation sinks. The biogenic emissions were calculated online by using the Model of Emissions of Gases and Aerosols from Nature (MEGAN), which is embedded in WRF-Chem (Guenther et al., 2006). This module estimates the net emission rates of monoterpene, isoprene, and other BVOCs from terrestrial ecosystems to the above-canopy atmosphere.

The key parameterization options applied in this study include the Yonsei University (YSU) boundary layer scheme (Hong et al., 2006), the Monin–Obukhov surface layer scheme (Jimenez et al., 2012), the unified Noah land surface scheme (Ek et al., 2003), the Morrison microphysics scheme (Morrison et al., 2009), the Grell–Freitas cumulus parameterization scheme (Grell and Freitas, 2014), and the Rapid Radiative Transfer Model for General Circulation Models (RRTMG) longwave and shortwave radiation (Iacono et al., 2008). For simulating gas-phase chemistry, the Statewide Air Pollution Research Center (SAPRC-99) mechanism developed by Carter (2000) was employed. Aerosol lifecycle processes were represented by the updated Model for Simulating Aerosol Interactions and Chemistry (MOSAIC) module, which utilizes a sectional framework with 20 logarithmically spaced bins ranging from 1 nm to 10 µm. This sectional approach enables better representations of nucleation and particle growth processes (Lupascu et al., 2015; Matsui et al., 2011; Lai et al., 2022a, b). The condensed volatility basis set (VBS) mechanism proposed by Shrivastava et al. (2011) was used to represent the gas–particle partitioning and multi-generational gas-phase oxidation of organic vapors, with additional updates for biogenic secondary organic aerosol (SOA) formation (Schervish and Donahue, 2020; Shrivastava et al., 2015).

We evaluated the performance of our model simulations by comparing them with meteorological parameters such as 2 m temperature and 10 m wind speed, as well as NPF-related pollutants such as SO₂ and PM_2.5 measurements collected in Lulang. The results demonstrate that our model is capable of reproducing meteorological conditions and NPF-relevant pollutants well around the Lulang site (Table 1). Furthermore, our model can appropriately capture the regional spatial distributions of meteorological factors and NPF-relevant pollutants (Fig. S1). To investigate the role of NPF and the specific role of organic NPF pathways, we conducted three parallel numerical experiments: (1) a simulation with all NPF pathways turned off (NPF-off), (2) a simulation considering all NPF pathways (NPF-on), and (3) a simulation including only inorganic NPF pathways (NPF_inorg).

Table 1 Statistical analysis of the simulated 2 m temperature, 10 m wind speed, SO₂ concentration, and PM_2.5 concentration versus the observations in Lulang.

^∗ MB and RMSE refer to mean bias and root mean square error, respectively.

Figure 2 (a) Number concentration of particles larger than 50 nm (NC_>50 nm) as a function of black carbon concentration [BC] in Lulang during the campaign. The lines represent those fittings for S1 in which 0.1 %, 0.5 %, 1 %, 1.5 %, 2 %, or 5 % of the data points are located below the line represented by S1 × [BC]. (b) Diurnal variation in the fraction of secondary source contributions to NC_>50 nm in Lulang during the campaign.

2.3 Improvements in the new particle formation scheme in WRF-Chem

In this study, we employed a modified version of the VBS approach to calculate the organics contributing to NPF (Schervish and Donahue, 2020; Shrivastava et al., 2015). The default VBS mechanism represents non-traditional SOA derived from the oxidation of semi- and intermediate-volatility primary gases by two volatility species with saturation mass concentration (C^∗) values of 10⁻² and 10⁵ µg m⁻³ (at 298 K and 1 atm) while treating traditional SOA produced by oxidation of biogenic and anthropogenic VOCs with C^∗ values of 10⁰ µg m⁻³. Previous studies have demonstrated that the oxidation of monoterpene could generate large amounts of ultra- and extremely low-volatility organic compounds (ULVOCs and ELVOCs), defined as organics with C^∗ values smaller than $\mathrm{3}\times {\mathrm{10}}^{-\mathrm{9}}$ µg m⁻³ and $\mathrm{3}\times {\mathrm{10}}^{-\mathrm{5}}$ µg m⁻³, respectively, which could efficiently contribute to nucleation and the initial growth process (Ehn et al., 2014; Zhao et al., 2020). Importantly, the formation of ULVOCs and ELVOCs occurs through autoxidation and dimerization processes involving peroxy radicals (RO₂), which differ from the traditional oxidation reactions commonly considered in the models (Crounse et al., 2013). To accurately depict the formation of low-volatility organic compounds from monoterpene, we extended the VBS mechanism to explicitly incorporate the chemistry of RO₂, including autoxidation, dimerization processes, and other competing reaction pathways. The resulting stable molecules from the termination of RO₂ radicals were then distributed into the appropriate volatility bins (Schervish and Donahue, 2020). The details about chemical processes and key parameters within the modified VBS are described in the Supplement (Sect. S2). To account for the involvement of biogenic precursors in NPF, we adjusted the volatility representation of biogenic species within the VBS mechanism. Specifically, we reduced the lowest-volatility bin in the mechanism from 10⁰ to 10⁻⁹ µg m⁻³, allowing for the representation of ULVOCs and ELVOCs. Additionally, we expanded the volatility distribution to include eight bins with C^∗ values ranging from 10⁻⁹ to 10⁵ µg m⁻³, separated by 2 orders of magnitude, to adequately capture the wide volatility range of organic vapors.

To simulate NPF processes, we incorporated a comprehensive ensemble of eight NPF parameterizations based on laboratory measurements into the MOSAIC module, coupled with the improved VBS scheme described above. These parameterizations accounted for both inorganic and organic nucleation pathways, including binary-neutral and ion-induced NPF (H₂SO₄–H₂O); ternary-neutral and ion-induced NPF (H₂SO₄–NH₃–H₂O); amine–H₂SO₄ NPF (H₂SO₄–amine–H₂O); and three organic pathways, including multicomponent organic NPF (H₂SO₄–NH₃–organics–H₂O), pure-organic neutral NPF, and pure-organic ion-induced NPF. The inorganic nucleation pathways are based on Dunne et al. (2016) and have been described in detail in our previous study (Lai et al., 2022b). As for the organics-involved nucleation, parameterizations of H₂SO₄–NH₃–organics–H₂O (Lehtipalo et al., 2018) and pure organics (Kirkby et al., 2016) were adopted. Importantly, Lehtipalo et al. (2018) have demonstrated that out of all HOMs, only organic vapors with ultra-low or extremely low volatility are able to participate in nucleation. Thus, species with C^* values lower than 10⁻⁷ µg m⁻³ were considered the nucleating vapors in our model. It should be noted that, as for the pure-organic NPF, the original parameterization utilized the measured HOMs with a wide volatility range ($C^{*}={\mathrm{10}}^{-\mathrm{20}}$–10² µg m⁻³) as input variables, whereas we opted for species with C^* values lower than 10⁻⁷ µg m⁻³. To account for the changes in the input variables and maintain consistency, we introduced a linear adjustment factor of 6, which was determined as the ratio of the HOMs with $C^{*}<{\mathrm{10}}^{-\mathrm{7}}$ µg m⁻³ to total HOMs measured in the laboratory study (Kirkby et al., 2016). Moreover, a temperature dependence function following Dunne et al. (2016) was introduced to account for the temperature effects on H₂SO₄–NH₃–organics–H₂O nucleation (details about the temperature dependence can be found in the Supplement; Sect. S3).

3 Results and discussions

3.1 Observational evidence for the enhanced NPF by anthropogenic and biogenic interactions

As a relatively remote natural area, NPF could be one of the main sources of CCN on the southeastern TP. We employed the method proposed by Kulmala et al. (2016) to estimate the source of CCN (details about the method can be found in the Supplement; Sect. S4). The particles with a size larger than 50 nm were used as the CCN proxy. Figure 2a illustrates the determination of the semi-empirical scaling factor (S1), which is crucial for estimating the relative contributions of primary and secondary sources to CCN. To evaluate the sensitivity of the estimation to S1, six different criteria were used. Specifically, the scatter plot was constrained such that 0.1 %, 0.5 %, 1 %, 1.5 %, 2 %, or 5 % of the data points fell below the line represented by S1 × [BC]. Previous studies using the same method have reported S1 values ranging from 3×10⁶ to 10×10⁶ CCN per nanogram of BC (Rodriguez and Cuevas, 2007; Fernández-Camacho et al., 2010; Gonzalez et al., 2011). However, our study found that the S1 values for CCN in Lulang ranged between 0.15×10⁶ and 0.60×10⁶ CCN per nanogram of BC, which are considerably smaller than those reported earlier. Figure 2b presents the average fraction of secondary particles in the CCN under different constraints for determining S1, revealing variations between 62 % and 90 %. Despite the associated uncertainty, our findings suggest that secondary sources, such as NPF, are the dominant contributors to the CCN budget in Lulang.

Figure 3 Temporal evolutions of (a) particle number size distribution from 1 to 1000 nm and (b) positive ion number size distribution from 0.8 to 40 nm. Note that the high peaks of number concentrations around 10–200 nm observed in the morning were caused by the wood burning in a single residential cottage near the site. (c) SO₂, monoterpene, and O₃ concentrations measured at the Lulang site during 28–30 April 2021. All times are in the UTC+8 time zone in this study.

The statistical results of NPF events during the campaign are given in Table S1. Notably, more than 60 % of clear-sky days were NPF event days, according to the criteria proposed by Dal Maso et al. (2005). The time series of particle number size distribution from 1 to 1000 nm, ion number size distribution, SO₂, monoterpene, O₃, H₂SO₄, and HOMs during the whole campaign are presented in Fig. S2. Figure 3a and b display the temporal variations in particle and ion number size distributions during a clear-sky episode from 28 to 30 April 2021, respectively. Throughout this episode, typical NPF events were observed on a daily basis. The temporal variation in particle number size distribution showing a typical “banana shape” (Fig. 3a) suggests that nucleation and growth occur on a regional scale (Kulmala et al., 2012). Meanwhile, the local wood burning originating from the cottage mentioned had minimal influence on NPF (Fig. S3), reinforcing the notion that our results can be considered applicable to the regional scale. Figure 3c illustrates the concentrations of NPF-relevant trace gases. The concentration of SO₂ in Lulang remained below 0.3 ppb, potentially approaching the lower detection limits. O₃ concentration displayed pronounced diurnal variation, with maximum concentrations exceeding 50 ppb during the daytime (12:00–18:00, UTC+8), indicating active photochemical reactions and a strong atmospheric oxidizing capacity. Due to the faster reactions with oxidants and enhanced boundary layer mixing in the daytime, monoterpene, on the other hand, showed a diurnal variation with higher mixing ratios at night and lower concentrations during the day. Specifically, the daily mean concentration of monoterpene was 118 ppt, with daytime mixing ratios ranging from 20 to 200 ppt and nighttime mixing ratios exceeding 0.26 ppb. This monoterpene concentration is comparable to the measurements in a boreal forest of southern Finland (Hyytiälä), where the average daily mixing ratio was 61 ppt during the spring season (Hakola et al., 2012).

Figure 4 (a) Nucleation rates (J_1.7) as a function of H₂SO₄ concentration at ambient observations in Lulang (green circles), in Hyytiälä (gray circles) (Sihto et al., 2006; Kulmala et al., 2013), and for CLOUD experiments (red diamonds) (Lehtipalo et al., 2018). The cyan and blue lines denote ternary (H₂SO₄–NH₃–H₂O) nucleation and binary nucleation (H₂SO₄–H₂O), respectively, based on CLOUD data in Kürten et al. (2016). (b) Averaged diurnal variations in H₂SO₄ concentrations and HOM concentration on NPF days in Lulang. The lines are the median values, and shaded areas denote the 25th or 75th percentiles. (c) A mass defect plot illustrating the chemical composition of negative ion clusters at 12:00 (UTC+8) on 29 April. The size and color of symbol size correspond to the relative signal intensity on a logarithmic scale and carbon number, respectively. (d) Formation rate at 10 nm (J₁₀) versus formation rate at 3 nm (J₃) in Lulang. Diamonds are color-coded by condensation sink. Error bars present the 25–75th percentiles. The solid grey line shows the relationship between J₁₀ and J₃ based on theory (Kulmala et al., 2012), and the uncertainties are shown by the shaded bands. Dash 1:1 line is shown for reference.

Figure 5 (a) Simulated particle number size distribution in Lulang. (b) Observed and simulated H₂SO₄ and monoterpene concentration in Lulang. The horizontal lines of the box represent the 66th percentile, median, and 33rd percentile, and the whiskers represent the 75th and 25th percentiles. (c) Same as Fig. 5b but for number concentrations of 1–10 and 10–40 nm particles. (d) The observed and simulated particle number size distributions averaged from 12:00 to 18:00 (UTC+8) during 28–30 April. The gray squares and gray circles show the measurements by NAIS and SMPS, respectively. Red, blue, and orange lines represent the NPF-on, NPF_inorg, and NPF-off experiments.

In Fig. 4a, the measured nucleation rates in Lulang are compared with laboratory experiments in the CLOUD chamber (Kürten et al., 2016; Lehtipalo et al., 2018) and ambient observations in the boreal forest in Hyytiälä (Sihto et al., 2006; Kulmala et al., 2013). The HOMs produced by the monoterpene oxidation have been demonstrated to play an important role in NPF in the boreal forest (Rose et al., 2018; Lehtipalo et al., 2018; Mohr et al., 2019). As shown in Fig. 4a, the nucleation rate and H₂SO₄ concentration in Lulang are similar to the observations in Hyytiälä, indicating a similarity in the nucleation process between the two locations. The nucleation exhibits a weak H₂SO₄ dependency and cannot be explained by H₂SO₄–H₂O and H₂SO₄–NH₃–H₂O nucleation. However, the ambient atmospheric conditions in Lulang can be reproduced by the experiments considering three precursors (monoterpene, H₂SO₄, and NH₃), suggesting the significant role of the multicomponent acid–base–organic nucleation mechanism (H₂SO₄–NH₃–organics–H₂O).

The interactions between precursors from anthropogenic sources and HOMs produced from biogenic VOCs could dominate the NPF process in Lulang. As shown in Fig. 4b, both H₂SO₄ and HOM concentrations in Lulang exhibit pronounced diurnal cycles, with peaks in the afternoon due to higher oxidation capacity during the daytime. The median daytime concentrations of H₂SO₄ and highly oxygenated organic molecule (HOM) concentration are about 10⁶ cm⁻³ and 2×10⁸ cm⁻³, respectively, consistent with measurements conducted in other forest environments (Riccobono et al., 2014). Figure 4c shows mass defect plots for negative ion clusters measured by APi-TOF during NPF on 29 April. Many ions with $m/z$ values higher than 300 Th and carbon number higher than 10 were observed during the NPF, suggesting the contribution of HOMs to the nucleation. The average mass spectrum of neutral molecules during 28–30 April 2021 is presented in Fig. S4. The spectrum displayed the patterns of HOM monomer and dimer from monoterpene oxidation. The C5–10 HOMs dominated in terms of both species number and concentration. The median double-bond equivalent (DBE) and DBE per carbon of neutral HOMs were 2.97 and 0.34, respectively. The C10 HOMs dominated the carbon distribution of HOM concentrations, with a fraction of 32 %. Moreover, a certain concentration of HOM dimer was observed. All this evidence suggests that the HOMs are mainly formed by the oxidation of monoterpene oxidation.

Figure 6 (a) Averaged spatial distributions of simulated near-surface 1–10 nm particle number concentrations during 28–30 April. Note that the black star marks the location of the Lulang site. The dashed pink rectangle defines the further research domain. (b) The relative contribution of different NPF pathways to the near-surface nucleation rate averaged over the region indicated by the dashed pink rectangle in Fig. 6a during 28–30 April.

In Fig. 4d, we compared the observed particle formation rates at 3 and 10 nm in Lulang, which are calculated based on NAIS measurement. According to the classical theory that neglects the influence of transport, it is expected that J₁₀ would be much lower than J₃ due to particle losses through coagulation scavenging as particles grow larger (Kerminen et al., 2001; McMurry et al., 2005; Kulmala et al., 2012). However, the observed J₁₀ is higher than what is predicted by the theory on most NPF days and notably higher than the observed J₃ on 29–30 April and 21–22 May. One possible explanation is that the sub-10 nm particles were formed elsewhere and transported to the measurement site. A previous study also reported the significant role of transport, whether horizontally or vertically from upwind regions, in the NPF events in Lulang (Cai et al., 2018). The potential role of transport is also illustrated in Fig. 3a, which shows the growth but not cluster formation. These findings suggest a non-negligible influence of transport on the observed NPF event.

3.2 Simulations on H₂SO₄–NH₃–organics–H₂O nucleation on the southeastern TP

To further quantify the roles of BVOCs and transport in the NPF in Lulang, we conducted regional simulations using the updated WRF-Chem regional atmospheric transport model for a typical NPF episode (28 to 30 April 2021). Figure 5a demonstrates that the model, incorporating a comprehensive ensemble of NPF pathways, was able to reproduce the occurrence of NPF events on these days. In addition, we compared the simulated concentrations of key NPF precursors with observations (Fig. 5b). The results showed that the model was able to reproduce the magnitude of the precursor concentrations, although it tended to underestimate the monoterpene concentration with a normalized mean bias of −16 %. This could be attributed to the simplification of vegetation classification and the limited understanding of the emission factors in the numerical descriptions (Guenther, 2013). To validate our simulation results, we investigated the behavior of particle number concentrations in the range of 1–10 nm (CN₁₋₁₀) and 10–40 nm (CN₁₀₋₄₀). As shown in Fig. 5c, the simulated CN₁₋₁₀ values are quite close to observations, while CN₁₀₋₄₀ tended to be overestimated. The discrepancy of particle number concentrations in the large size range could be caused by several factors such as emission inventories, topography, and model resolution. We compared the simulated particle number size distributions with the observations (Fig. 5d). The inclusion of inorganic NPF (“NPF_inorg”) narrowed the gap between simulation and observation but still could not reproduce the observed peak of small particles (< 40 nm diameter). The further incorporation of organic NPF pathways (NPF-on) bridged the remaining gap and resulted in a fairly good agreement with observations, indicating the vital role of organics in nucleation.

Figure 7 Averaged spatial distributions of simulated near-surface (a) SO₂ concentration, (b) OH concentration, (c) condensation sink, and (d) H₂SO₄ concentration during 28–30 April. Note that the black star marks the location of the Lulang site.

Figure 8 (a) Spatial distributions of simulated near-surface NH₃ concentration together with surface wind fields averaged from 12:00 to 18:00 (UTC+8) during 28–30 April. (b) Latitude–height cross sections of NH₃ concentrations together with wind fields averaged along 94–96° E from 12:00 to 18:00 during 28–30 April. Note that the dashed pink lines show the location of the research domain marked in Fig. 6a. (c) Averaged spatial distributions of simulated near-surface monoterpene concentration during 28–30 April. (d) Same as Fig. 8c but for oxidized biogenic gas. Note that the black star marks the location of the Lulang site.

Figure 6a presents a simulated spatial distribution of near-surface CN₁₋₁₀ averaged during the episode, revealing high concentrations of CN₁₋₁₀ on the southeastern TP. Our model results indicate that the nucleation and growth of aerosol particles extended over a large regional area, covering several thousand kilometers of the southeastern TP. To gain a deeper understanding of the nucleation and growth process on the southeastern TP, we extend the analysis from Lulang to the NPF-intensive surrounding area, as marked by the dashed pink rectangle in Fig. 6a. Figure 6b illustrates the relative contribution of different NPF pathways within this area. Our findings indicate that the pathways involving organic compounds dominate the NPF rates on the southeastern TP, with H₂SO₄–NH₃–organics–H₂O nucleation and pure-organic ion-induced NPF accounting for 85.0 % and 12.6 %, respectively.

Figure 9 (a) Diurnal evolution of contributions of near-surface 1–10 nm particle number concentrations from different processes calculated by WRF-Chem process analysis during 28–30 April over the research domain. The dashed black line represents the diurnal evolution of observed 1–10 nm particle number concentrations during 28–30 April. (b) Same as Fig. 9a but for 10–40 nm particle number concentrations.

To investigate the sources of gaseous precursors of H₂SO₄–NH₃–organics–H₂O nucleation on the southeastern TP, we present simulated spatial distributions of SO₂, OH, H₂SO₄ concentration, and condensation sink (CS) at the ground surface during the NPF episode (Fig. 7). SO₂, primarily derived from coal combustion in power generation and industrial production, displays high concentrations in South Asia (Fig. 7a). In contrast, simulated SO₂ concentration over the TP was less than 0.5 ppb, which is supported by our observations (Fig. 7a). The OH radical concentration, which is a key oxidant for forming H₂SO₄ and an indicator of atmospheric oxidizing capacity, is high on the southeastern TP (Fig. 7b). This high concentration is consistent with previous studies and can be attributed to strong solar radiation, high-ozone background, abundant water vapor, and low anthropogenic VOCs on the southeastern TP (Fig. S5) (Ren et al., 1999; Lin et al., 2008; Yan et al., 1997; Zhu et al., 2006; Wu et al., 2005; Wang et al., 2015). CS, which quantifies the role of pre-existing particles in removing condensable vapors and newly formed particles from the atmosphere, is expected to suppress NPF (Kulmala et al., 2012; Kerminen et al., 2018; Qi et al., 2019, 2015). Due to the relatively pristine environment, lower CS values are found over the TP compared with those in South Asia (Fig. 7c). As a result of strong atmospheric oxidizing capacity and low CS, H₂SO₄ concentrations can reach relatively high values on the TP compared to the surrounding area despite the low SO₂ concentration (Fig. 7d).

Figure 10 (a) Averaged diurnal evolution of vertical cross section of simulated nucleation rate (J_1.7) over the research domain during 28–30 April. Note that the dashed black line shows the planetary boundary layer height (PBLH). (b) The relative contribution of different NPF pathways averaged over the research domain during 28–30 April. (c) Same as Fig. 10a but for 10–40 nm particle number concentrations. (d) Same as Fig. 10a but for 10–40 nm biogenic organic mass concentrations.

Ammonia (NH₃) is also a significant precursor for particle formation, mainly emitted from agricultural, domestic, and industrial activities (Behera et al., 2013). Given the large footprint of agriculture in this region, the South Asia region, located to the south of the TP, has been identified as a hotspot for ammonia emissions (Clarisse et al., 2009; Van Damme et al., 2018; Xu et al., 2018b). Figure 8a shows that high levels of near-surface NH₃ were simulated in South Asia. The vertical and horizontal simulations of wind field, as well as the spatial distribution of simulated NH₃ concentrations during the daytime (Figs. 8a and b and S6), suggest that an overall northward circulation and up-valley winds favor the transport of NH₃ from South Asia to the TP. This transport of anthropogenic pollutants to the TP contributes to considerable NH₃ and SO₂ concentrations on the southeastern TP, facilitating H₂SO₄–NH₃–organics–H₂O nucleation. Moreover, the southeastern TP is characterized by high monoterpene concentrations due to the extensive coverage of evergreen forests and high monoterpene emission (Figs. 1a and 8c). Given the substantial concentration of monoterpene and the strong oxidizing capacity in the region, abundant oxidized biogenic gases are formed on the southeastern TP (Fig. 8d), which could significantly contribute to the NPF (Ehn et al., 2014; Bianchi et al., 2016; Qi et al., 2018). In summary, NPF is promoted on the southeastern TP through anthropogenic–biogenic interactions.

Figure 11 Averaged diurnal evolution of vertical cross section of (a) vertical mixing tendencies for 1–10 nm particle number concentrations, (b) condensation tendencies for 10–40 nm particle number concentrations, (c) vertical mixing tendencies for 10–40 nm particle number concentrations during 28–30 April over the research domain. Note that the dashed black line denotes PBLH.

3.3 The vertical heterogeneity of NPF on the southeastern TP

The observational evidence presented in Sect. 3.1 as well as Fig. 4d suggests that the observed NPF was highly likely to be influenced by regional transport and convective motions. To further investigate the impacts of different chemical and physical processes on NPF, we conducted the diagnostic analysis using the WRF-Chem model. WRF-Chem demonstrates good agreement in reproducing the diurnal variation and magnitude of particle number concentrations during the daytime (Fig. S7). Our focus was specifically on CN₁₋₁₀ and CN₁₀₋₄₀, which serve as indicators of J₃ and J₁₀, respectively. Figure 9 illustrates the diurnal evolution of the relative contributions from nucleation, condensation/evaporation, coagulation, and transport (vertical mixing and advection) to near-surface CN₁₋₁₀ and CN₁₀₋₄₀, along with the observed CN₁₋₁₀ and CN₁₀₋₄₀. CN₁₋₁₀ exhibits a dominant peak around 14:00, which is closely associated with the nucleation process (Fig. 9a). The main loss of CN₁₋₁₀ is coagulation and vertical mixing. As for CN₁₀₋₄₀, a peak is observed around 17:00, with a time lag of 3 h compared to CN₁₋₁₀. The diagnostic analysis reveals that vertical mixing is the dominant process that enhances the surface CN₁₀₋₄₀ during the daytime, with an estimated contribution of 272 % (Fig. 9b). The condensation process shows a negative contribution to CN₁₀₋₄₀ during the daytime and a prominent positive contribution in the evening, indicating that the particles shrink and grow through evaporation and condensation, respectively. Similar to CN₁₋₁₀, coagulation serves as one of the main losses for CN₁₀₋₄₀. In summary, our diagnostic analysis reveals that nucleation predominantly drives the production of near-ground CN₁₋₁₀, while vertical mixing enhances the concentration of particles in the larger size range near the ground.

To gain further insights into the vertical distribution of the NPF process, we analyzed the simulated vertical profiles of nucleation rate during the NPF episode over the research domain (Fig. 10a). The nucleation rate exhibits a distinct vertical pattern, with peak values observed near the surface within the planetary boundary layer (PBL), as well as another peak at around 5 km in the free troposphere. This vertical distribution suggests that nucleation not only occurs predominantly in the PBL but also extends to higher altitudes in the atmosphere. We further examined the relative importance of each nucleation mechanism (Fig. 10b). At lower altitudes, H₂SO₄–NH₃–organics–H₂O nucleation and organic ion-induced nucleation dominate the nucleation rate, accounting for 75 % and 12 % below 2 km, respectively. As the altitude increases, H₂SO₄–NH₃–H₂O nucleation becomes the primary pathway between 4 and 5 km, with ion-induced and neutral pathways contributing 12 % and 83 %, respectively. NH₃ exhibits higher concentrations near the surface within the PBL and another peak at around 5 km (Fig. S8a in the Supplement). The higher concentrations of NH₃ at lower altitudes are attributed to transport from South Asia by the near-surface southerly winds (Fig. 8a). The NH₃ could then be transported into the upper troposphere through convection over the TP. The presence of elevated NH₃ concentrations in the free troposphere has also been reported in previous studies based on aircraft observations (Höpfner et al., 2016, 2019).

After nucleation, the newly formed particles undergo growth through condensation of vapors onto their surface, resulting in an increase in particle size. Interestingly, in contrast to the vertical distribution of nucleation rate, the CN₁₀₋₄₀ exhibited a notable counter gradient during the daytime (Fig. 10c). This pattern is consistent with the distribution of biogenic organic mass concentration (Fig. 10d), indicating the important role of biogenic organic vapors in this region. The lower PM_2.5 concentration at the boundary layer top suggests a lower condensation sink for condensable gases and a higher survival probability of freshly nucleated particles (Fig. S8b). Moreover, as temperature decreases with increasing altitude, the volatility of organics decreases, although the autoxidation process is suppressed (Stolzenburg et al., 2018). Consequently, the lower temperature at higher altitudes creates more favorable conditions for the condensation process of organic vapors, facilitating the subsequent growth of nanoparticles (Fig. S8c).

Figure 11 further elucidates the key processes in shaping the vertical heterogeneity of NPF. As discussed in Sect. 3.2, the near-surface nucleation induced by the anthropogenic–biogenic interaction is facilitated over the southeastern TP, leading to the formation of substantial nanoparticles. These newly formed nanoparticles are then transported upwards through vertical mixing within the PBL (Fig. 11a). As these particles ascend, they encounter a notable reduction in air temperature and PM_2.5 concentration in the upper PBL (Fig. S8b, c). These altered conditions create an environment conducive to the condensational growth of small particles (Fig. 11b), resulting in an increased concentration of particles within the 10–40 nm size range in the upper air (Fig. 10c). Subsequently, the particles that underwent growth through condensation in the upper air are effectively brought back down to lower altitudes by vertical mixing (Fig. 11c). This process contributes to a significant enhancement of particle concentration in the size range of 10–40 nm near the surface, as observed in the CN₁₀₋₄₀ measurements (Fig. 9b). These findings highlight the importance of vertical mixing in the vertical redistribution of particles and provide valuable insights into the complex dynamics of NPF on the southeastern TP.

4 Conclusions

Combining comprehensive in situ measurements and an updated regional chemical transport model, this study provides a comprehensive investigation of the NPF occurring on the southeastern TP during the pre-monsoon period. The measurement site, Lulang, exhibits regionally low anthropogenic emissions but a high-biogenic-emission environment, with trace-level concentrations of SO₂ and high concentrations of monoterpene. Our findings reveal the significant contribution of secondary sources to the CCN budget in Lulang, highlighting the importance of NPF in this region. On clear-sky days, a high frequency of NPF exceeding 60 % was observed. Observational evidence shows that the multicomponent organic (H₂SO₄–NH₃–organics–H₂O) mechanism is the dominant nucleation in Lulang, and HOMs formed from monoterpene oxidation play an important role. Model results further confirm that this nucleation mechanism prevails on the whole southeastern TP and reveals that NH₃ was transported from South Asia.

Importantly, our study also unveils the crucial role of vertically heterogeneous NPF and vertical mixing in shaping the particle number size distribution near the surface. The vertical distribution of precursors (such as monoterpene and NH₃) leads to strong nucleation near the ground, and newly formed small particles are transported upward. The upper PBL exhibits lower temperature and PM_2.5, favorable for condensational growth, resulting in a counter-gradient pattern for CN₁₀₋₄₀ during the daytime. As the PBL evolves, larger particles are transported to the ground, leading to a significant enhancement of CN₁₀₋₄₀ near the surface. These processes, including the anthropogenic–biogenic NPF and meteorological dynamics, can explain the observed high ratio of J₁₀ to J₃ during the NPF on the southeastern TP.

This study gives valuable insights into the prevailing mechanism driving NPF on the southeastern TP, highlighting the interaction between anthropogenic and biogenic sources in NPF processes in this region. We emphasize the importance of considering vertical mixing to improve our understanding of NPF and its implications for climate change. Additionally, our findings underscore the significant role of NPF at high altitudes, particularly in pristine regions like the TP, as they can profoundly influence cloud properties and have broader implications for climate dynamics.