Introduction

Against the backdrop of the prevailing global climate change, the investigation of aerosol activation has assumed heightened importance. Serving as the pivotal link between aerosols and clouds, aerosol activation plays a pivotal role in the intricate interactions between aerosol and cloud1,2, thereby exerting profound influences on precipitation3,4, radiation5,6,7,8, and the broader global climate system9,10. The latest assessment report from the Intergovernmental Panel on Climate Change (IPCC) underscores that aerosol-cloud interactions remain the most significant source of uncertainty in climate predictions11. Consequently, research endeavors focused on unraveling the intricacies of aerosol activation have emerged as a critical pathway to mitigate this uncertainty. The imperative to incorporate aerosol activation treatments into climate models has been accentuated, with the goal of refining predictions related to cloud properties and radiative forcing12,13.

The theoretical basis of the aerosol activation process is rooted in Köhler theory14, which enables the prediction of particle size growth for dry aerosols with known physical and chemical properties under specific environmental humidity conditions. However, its application to the natural atmosphere is not straightforward. This complexity arises from the intricate characterization of aerosol chemical properties in the Raoult term, necessitating input parameters like the van’t Hoff factor, solute mass, molar mass, and others14. The diverse chemical composition of aerosols in the natural atmosphere further complicates obtaining these quantities. Consequently, early parameterizations of aerosol activation relied heavily on empirical fitting relationships between cloud condensation nuclei (CCN) number concentration (NCCN) and the vapor supersaturation (SS), e.g., the classical power-law equation15. Petters and Kreidenweis16 introduced a single parameter, aerosol hygroscopicity (κ), which relates the volume of water taken up by a particle to the water activity. This significantly improved the convenience of the Köhler theory, as it parameterizes the aerosol chemical properties in the Raoult term using a single variable. This advancement has led to a novel approach in aerosol activation parameterization, integrating the dry particle number size distribution (PNSD) with a parameterized κ to calculate NCCN based on the κ-Köhler theory e.g.,17,18,19. This, in turn, motivated observational experiments on aerosol κ globally, aiming to capture the characteristics of κ in various regions, such as the Amazon rainforest20,21, urban and suburban areas22,23,24,25, rural locations26,27, coastal regions28,29, and high-altitude stations30,31. Summarizing various field experiments, Andreae and Rosenfeld32 proposed κ values of 0.3 ± 0.1 and 0.7 ± 0.2 as representative for continental and marine aerosols, respectively. Subsequent studies have corroborated similar ranges of κ distributions and utilized this simple κ parameterization for activation calculations e.g.,22,27,33,34,35,36.

The Tibetan Plateau (TP), considered the “Roof of the World”, the “Asian Water Tower”, and the “Heat Source Pump”, wields a profound influence on regional precipitation and even global climate dynamics37,38. Given its distinctive geographic and climatic conditions, the exploration of aerosol, cloud, and precipitation processes in this region is crucial yet faces challenges. Especially, the κ and activation characteristics of aerosols in the TP region may deviate from the normal continental conditions, which, however, remain largely unexplored. To our knowledge, only Xu et al.39 reported the NCCN(SS) measured at the eastern TP sites. The specific characteristics of κ in the TP region have yet to be documented. This knowledge gap does not stem from the insignificance of aerosol activation properties in the TP region but rather arises from observational challenges. Additionally, the unique topographical conditions, variability in aerosol sourcing, and the intricate interplay between topography and atmospheric flows collectively obscure a comprehensive understanding of the impacts on cloud and precipitation processes in the TP region40,41,42. Unraveling these complexities necessitates a foundation of high-precision, in-situ aerosol-cloud-precipitation microphysical observations. However, the formidable challenges linked to in-situ observations in the TP region hinder the implementation of such measurements. Consequently, current researches on aerosol-cloud-precipitation in this region mainly rely on satellite remote sensing and model simulations e.g.,42,43,44,45, with scant reports of in-situ observational outcomes.

To advance our comprehension of aerosol-cloud-precipitation microphysical characteristics and their interactions within the Southern Tibetan Plateau (STP) region, we carried out the Ground in-situ Aerosol-Cloud-Precipitation Experiment on the STP (GACPE-STP) from August 17th to October 18th, 2023. As shown in Fig. 1, the research site (27.42 N, 88.90 E, 3136 m above sea level) is located in the mountainous region of Yadong County, China, in close proximity to the Indian border. The method section provides an overview of this endeavor. This study focuses on the characteristics of aerosol activation with three principal objectives: firstly, to elucidate the overall κ and activation characteristics of aerosols in the region; secondly, to access the appropriateness of the recommended continental κ value of 0.332 within the local context, evaluating its potential influence on predicting of cloud properties; and thirdly, to develop a κ parameterization for this region.

Fig. 1: Geographic location of the GACPE-STP measurement site.
figure 1

Colored curves represent three clusters of three-day back trajectories generated using the HYSPLIT model.

Results

Overall aerosol activation characteristics over STP

Figures 2a, b illustrate the temporal distribution of dry particle number size distribution (PNSD), total aerosol number concentration (Na), aerosol median diameter (Dmedian), and NCCN at various SS conditions during the observational period. The PNSD is predominantly characterized by an unimodal distribution, with a daily mean peak diameter varying in the range of 50 to 143 nm. The region exhibits relatively low Na, with a mean value of 962 cm−3. Daily mean Na varies considerably, reaching a maximum of 1671 cm−3, which is 4.3 times the minimum of 387 cm−3. The daily-mean Dmedian also displays significant variations, with a mean of 93 nm and a maximum of 136 nm, representing 2.5 times the minimum of 54 nm. Additionally, a notable positive correlation exists between Na and Dmedian (Supplementary Fig. 1), potentially linked to frequent cloud precipitation events during the observational period. Wet scavenging plays a pivotal role in reducing larger aerosol particles, simultaneously lowering Na and Dmedian.

Fig. 2: Time series of aerosol physical and activation properties and the CCN spectra.
figure 2

a Time series of daily-mean dry particle number size distribution and (b) dry aerosol number concentration (Na), median diameter (Dmedian), and CCN number concentration (NCCN) at various supersaturation (SS) conditions; c Relationships between NCCN and SS at Yadong and other stations. The error bar represents one standard deviation.

As SS increases, the critical dry activation diameter (Dc) decreases16, leading to an increase in NCCN. Unactivated particles significantly impact the CCN measurements, especially at low SS conditions46,47. At SS of 0.07%, the uncorrected daily-mean NCCN is 68 cm−3, corrected to only 24 cm−3, with unactivated particles contributing to 65% of the measured NCCN. At SS of 0.1% and 0.15%, the percentages decrease to 54 and 32%, respectively. Measurements at SS above 0.15% remain unaffected by unactivated particles. The region exhibits low NCCN and activated fraction (AF, i.e., NCCN/Na) and the daily-mean values increase from 24 cm−3 to 483 cm−3 and 2 to 48%, respectively, as SS increases from 0.07% to 0.7%. The relatively low AF corresponds to the large Dc at each specific SS (Supplementary Fig. 2), implying weak κ, a characteristic explored in detail in the subsequent section. Figure 2c depicts that, within the same SS range, the NCCN in the STP region is significantly lower than observations in rural e.g.,26,27 and urban areas e.g.,22,48, falling within the NCCN(SS) values observed in two Amazon tropical rainforest experiments20,21. The NCCN(SS) in the STP region is comparable with measurements recorded at the Qilian Observation and Research Station of Cryosphere and Ecologic Environment (LHG station) in the eastern TP region39. However, it stands distinctly lower than the NCCN(SS) obtained at the Waliguan station in the eastern TP region39. Notably, these two stations, the LHG and Waliguan stations, represent the only reported locations with NCCN(SS) measurements in the TP regions.

Furthermore, the positive correlation between Na and Dmedian increases the differences in NCCN at different SS, reflected in an elevated k value of power-law fit15. The power-law fit for corrected NCCN is represented as

$${N}_{{CCN}}=887{{SS}}^{1.25}$$
(1)

with R2 of 0.92, where the k value of 1.25 is notably higher than the characteristic range of 0.4–0.9 summarized for continental regions by Seinfeld and Pandis49. The fitting results of the corrected NCCN(SS) using other functions are provided in Supplementary Fig. 3.

Various air masses exhibit noteworthy variations in aerosol activation characteristics. The prevailing air masses predominantly originated from the southern regions with relatively lower altitudes based on the simulations from HYSPLIT model, as shown in Fig. 1. According to the distance of air mass transport, it can be classified into three categories, as detailed in the Method Section. In Fig. 3, when air masses originate from local emissions and short-distance transport, designated as C1, the PNSD is characterized by the highest number concentration and largest particle size, with maximum values for Na and Dmedian of 1091 cm−3 and 101 nm, respectively. Although κ is not the largest of the three clusters, the largest Na and Dmedian contribute to the largest NCCN. As the transport distance increases, the ability of aerosol activation weakens. In the scenario of air masses originating from long-distance transport, represented by C3, the PNSD becomes the lowest number concentration and smallest particle size, with the minimum values for Na and Dmedian of 642 cm−3 and 61 nm, respectively. Furthermore, the κ is also the lowest among the three air mass clusters. The Na, Dmedian, and κ are all the lowest, leading to the minimum NCCN among the three air mass clusters. A plausible explanation for this phenomenon is that air masses transported over long distances undergo uplift due to topography, resulting in cloud and precipitation formation, which substantially depletes the effective CCN.

Fig. 3: Physical and activation properties of aerosols from different air mass clusters.
figure 3

a The mean particle number size distribution (dNa/dlogD), b CCN number concentration at a supersaturation of 0.2% in column A (NCCN,0.2%,A), c total aerosol number concentration within a diameter ranging from 18.8 to 685.4 nm (Na), d median diameter of dry particles (Dmedian), and (e) hygroscopicity calculated by NCCN,0.2%,A (κccn,0.2%,A). The error bar represents one standard deviation. The trajectories of the air mass clusters are in Fig. 1.

Weak hygroscopicity and its size-dependent

In Fig. 4a, the temporal evolution of κ, calculated through various methodologies during the observational period, is presented. Generally, κ within the STP region is observed to be notably low. The mean values of κ, calculated from NCCN under a SS of 0.2% in column A (κccn,0.2%,A) and column B (κccn,0.2%,B) are 0.066 and 0.089, respectively. Additionally, the mean values of κ calculated from dry and wet particle median diameter, volume concentration, and scattering coefficient (κg, κv, and κf) are 0.054, 0.038, and 0.077, respectively. It is worth noting that κccn corresponds to hygroscopicity under supersaturated conditions, while κg, κv, and κf all represent hygroscopicity at RH < 100% (more often, less than 90%). Additionally, κf is calculated under set RH cycles, while the RH corresponding to κg and κv varies with ambient RH. Previous studies on hygroscopicity have found that using HTDMA (humidity tandem differential mobility analyzer) to measure hygroscopicity at RH < 100% and CCNC to measure hygroscopicity at RH > 100% yields different results e.g.,50,51. Additionally, Zhao et al.52 indicate that hygroscopicity varies with RH at RH < 100%. Therefore, the RH dependence of hygroscopicity could be a reason for the differences in hygroscopicity calculation results between these methods. Furthermore, the difference between κg and κv could be attributed to the neglect of curvature effects in the calculation of κv, while κg calculations do not neglect this term.

Fig. 4: Time series of various aerosol hygroscopicity (κ) values and their comparisons.
figure 4

a Time series of daily-mean κ values derived by NCCN in column A at SS of 0.2% (κccn,0.2%,A), NCCN in column B at SS of 0.2% (κccn,0.2%,B), dry and wet median diameter (κg), volume concentration (κv), and scattering coefficient of aerosol (κf); b Relationships among the five κ values; c Statistical mean value and one standard deviation of the five κ. The error bar represents one standard deviation.

These κ values consistently remain below 0.1, much lower than the mean (0.3) and even lower bound (0.2) of the recommended mean κ value for continental regions32. The low value of κ is intricately associated with the surroundings at the observational site. Notably, vegetation in the STP regions surrounding by forests and meadows is recognized for emitting volatile organic compounds (VOCs)53. The secondary organic aerosols (SOA) formed as precursors from VOCs exhibit either negligible or weak hygroscopicity. Examples include SOA formed as precursors from α-pinene and β-pinene, both exhibiting a reported κ value of 0.02254. Furthermore, inhabitants of the STP region adhere to distinctive lifestyles, utilizing domestic fuels that diverge from those commonly found in other regions of China. The combustion of biofuels such as yak dung, WeiSang mixture fuels, and powdery Tibetan incense, characterized by lower combustion efficiencies, leads to significant emissions of CO and organic aerosols55. Organic aerosols and black carbon, as the predominant components stemming from biomass burning activities, collectively contribute to more than 69.1−85.7% of the total PM1 mass at four sites in the southern and central TP39. These observations lend support to the measured low κ in the conducted experiment.

The daily-mean κccn displays a reduced temporal fluctuation when compared to κg, κv, and κf. This phenomenon can be ascribed to the relatively low aerosol loading at the measurement site. The calculation of κ, relying on measurements of both dry and wet aerosol properties (i.e., κg, κv, and κf), is notably influenced by instrumental bias due to the low aerosol loading, resulting in substantial fluctuations and occasional negative values. This effect further diminishes the correlation between the κ values calculated by different methods, as illustrated in Fig. 4b. The temporal trends of κccn,0.2%,A and κccn,0.2%,B exhibit similarity with a strong correlation. However, the absolute mean value of κccn,0.2%,B is elevated by 0.023 compared to κccn,0.2%,A, representing a 35% relative increase. This discrepancy is primarily attributed to the NCCN in column B being 24% higher than that in column A at SS of 0.2%.

Size-dependent κ can be calculated based on NCCN (SS) and dry PNSD. Figure 5 presents the relationships between κ and dry particle diameter (D) observed in this study, along with measurements from various regions. It is crucial to emphasize that under low SS conditions, the presence of unactivated particles in CCN measurements markedly increases the uncorrected NCCN, leading to a decrease in Dc and an increase in κ16. Consequently, a monotonically increasing relationship between κ and D is obtained. However, upon excluding unactivated particles, the authentic relationship between κ, calculated using corrected NCCN, and D conforms to a Gaussian distribution that κ initially increases with D and subsequently decreases as D increases. The fitting results are expressed as follows:

$${\kappa }_{{ccn}}=0.059+0.033\exp \left[-2{\left(\frac{D-159.8}{47.87}\right)}^{2}\right]$$
(2)
Fig. 5: Relationships between the aerosol hygroscopicity (κ) and dry diameter (D).
figure 5

The error bar represents one standard deviation. The black line is the Gaussian fit of the κ vs. D pairs observed in GACPE-STP, while the other lines represent the κ vs. D relationships observed at various stations.

with R² of 0.74. The peak of the fitting curve occurs at D of 159.8 nm, with a peak κ of 0.092. The κ(D) values in the STP region are consistently lower than those measured in rainforests e.g.,20,21, rural areas e.g.,26,27, and urban environments e.g.,22,56. This observation signifies a prevailing trend of low κ values for different particle sizes in the STP region.

Furthermore, owing to the SS cycling established in this experiment spanning from 0.07% to 0.7% and the low κ of the region, the mean Dc measured in this study is 391 nm at SS of 0.07%, decreasing to 92 nm at SS of 0.7%. Consequently, this experiment provides a relationship between κ and D in the mean D range of 92 to 391 nm. Notably, this range is more extensive than those obtained in previous CCN and HTDMA measurements, particularly in the larger particle size range. For instance, in Wang et al.27, a monotonically increasing relationship between κ and D was observed within the range of 40 to 200 nm. Additionally, in other studies e.g.,20,21,22,25,26, a consistent monotonically increasing relationship between κ and D was reported, but the D ranges in these studies were confined to 30 to 250 nm. Shen et al.56, employing an HTDMA to explore urban size-resolved κ across an extended size range of 50 to 600 nm, highlighted that κ initially increases with D and then decreases, deviating from a monotonically increasing trend. This behavior is ascribed to the lag time for different sizes to respond to the evolution of pollution. Concerning the κ-D relationship, the findings of Shen et al.56 align with our observations. Nevertheless, the mechanisms governing the variation of κ with D in the STP region warrant further investigation, particularly through the inclusion of size-resolved aerosol chemical composition.

Implications for aerosol activation and the aerosol indirect effects

Accurate prediction of cloud droplet number concentration (Nc) in weather and climate models hinges upon a comprehensive understanding of aerosol activation characteristics12. The approach based on the κ-Köhler theory utilizes parameterized κ and dry PNSD to predict NCCN, representing a significant development in aerosol activation parameterization. Its predictive efficacy notably surpasses that of empirical fits of NCCN and SS27. Consequently, an imprecise depiction of κ can introduce biases in predicting NCCN and Nc. When concerning aerosol activation and cloud formation, κccn emerges as the preferable parameter for characterizing κ, in contrast to κf, κg and κv. This preference arises because κccn embodies hygroscopicity under supersaturated conditions, reflecting the conditions under which aerosol activation and cloud formation occur. The relationship between κccn and D, derived from this observation and expressed in Eq. 2, can serve as a parameterization of κ for predicting NCCN in the STP region.

We evaluate deviations of calculating NCCN arising from employing κ of 0.3 and incorporating the fitted relationship κ(D) as outlined in Eq. 2. Figure 6 illustrates the relationship between measured NCCN and predicted NCCN under various SS conditions. Setting κ equal to 0.3 significantly overestimates NCCN in the STP region, particularly at low SS and low NCCN conditions, where the overestimation of NCCN can surpass one order of magnitude. The utilization of κ parameterized by Eq. 2 effectively mitigates this overestimation, resulting in improved predictions for NCCN in the STP region. It is noteworthy that, despite κ being set at 0.3 resulting in a significant overestimation of NCCN, there is a distinct positive correlation between the predicted and measured NCCN. This correlation is attributed to the use of measured dry PNSD in NCCN calculations, where particle size plays a more dominant role than chemical composition in predicting NCCN57. Figure 7a quantifies the measured and predicted mean NCCN under different SS conditions, which is then utilized to calculate the relative prediction biases shown in Fig. 7b. When employing κ = 0.3 to predict NCCN, a significant overestimation occurs, with a minimum bias of 77% at SS of 0.7% and a maximum bias up to 426% at SS of 0.1%. In contrast, using the κ(D) parameterization from Eq. 2 results in a shift in the mean NCCN prediction bias from −51 to 26% as SS increases from 0.07% to 0.7%, demonstrating better performance than assuming κ = 0.3. Therefore, the recommended κ = 0.3 for the continental regions32 does not hold for the STP region.

Fig. 6: Predicted vs. measured CCN number concentrations (NCCN).
figure 6

ai represent different supersaturation (SS) conditions (0.07% to 0.7%). Predictions are based on the κ-Köhler model approach using two types of effective hygroscopicity parameters (κ): κ vs. particle diameter (D) relationship in Eq. 2 (blue points) and constant value of κ = 0.3 (red points).

Fig. 7: The effect of weak hygroscopicity on aerosol activation and indirect effects.
figure 7

a Measured and predicted CCN number concentration (NCCN) at different supersaturation (SS) conditions; b Relative deviation between predicted and measured NCCN; c Relative deviation to cloud optical thickness (τ); d Relative deviation to the threshold function of the cloud‐to‐rain transition (Tauto).

The discrepancy in predicting NCCN directly translates to a forecast bias in Nc, subsequently impacting predictions of cloud radiative characteristics and precipitation processes, thereby influencing the assessment of aerosol indirect effects (AIEs). Lastly, we provide a brief evaluation of the impact of the two κ parameterization methods on AIEs. Cloud optical thickness (τc) can be approximately expressed as a function of cloud liquid water path (LWP) and cloud droplet effective radius (re) according to Stephens58:

$${\tau }_{c}\,\approx\, \frac{3{LWP}}{2{r}_{e}}$$
(3)

where re is proportional to volume-mean radius (rv):

$${r}_{e}=\beta {r}_{v}=\beta {\left(\frac{3{LWC}}{4\pi {\rho }_{w}{N}_{c}}\right)}^{1/3}$$
(4)

Here, LWC represents liquid water content, β is the ratio of re to rv, and is a function of relative dispersion of cloud droplet spectrum59. When neglecting dispersion effects, β can be assumed as a constant, e.g., β = 1.1 used in Quass et al.60. Additionally, Nc impacts the efficiency of cloud-rain autoconversion, thereby influencing the collision-coalescence and precipitation process. The formulation of the autoconversion threshold function (Tauto) is defined by Liu et al.61:

$${T}_{{auto}}=\left[\frac{{\int }_{{r}_{c}}^{{\infty }}{r}^{6}n\left(r\right){dr}}{{\int }_{0}^{{\infty }}{r}^{6}n\left(r\right){dr}}\right]\left[\frac{{\int }_{{r}_{c}}^{{\infty }}{r}^{3}n\left(r\right){dr}}{{\int }_{0}^{{\infty }}{r}^{3}n\left(r\right){dr}}\right]$$
(5)

where r is the cloud droplet radius, n(r) is the cloud droplet spectrum, and we utilize the n(r) observed by a fog monitor (FM-120, DMT Inc.) during a cloud event with drizzles on 2023/08/26 as an illustrative example for calculating Tauto. The critical radius for autoconversion (rc) is determined by an analytical expression based on Nc and LWC62:

$${r}_{c}\,\approx\, 4.09\times {10}^{-4}{\beta }_{{con}}^{1/6}\frac{{N}_{c}^{1/6}}{{{LWC}}^{1/3}}$$
(6)

where βcon is an empirical coefficient with a value of 1.15*1023 s−1. The LWC can be derived as:

$${LWC}=\frac{4\pi {\rho }_{w}}{3}{\int}_{\!0}^{{\infty}}{r}^{3}n\left(r\right){dr}$$
(7)

where ρw represents the liquid water density. Tauto ranges from 0 to 1, with higher values indicating an increased probability of collision-coalescence occurring in clouds.

In the assumption of constant LWC and LWP, an overestimation of Nc results in an underestimation of re, ultimately leading to an overestimation of τc. As shown in Fig. 7c, using κ = 0.3 causes an overestimation of τc ranging from 21 to 74% within the SS range of 0.07% to 0.7%. In contrast, when employing the κ(D) parameterization from Eq.2, τc is underestimated by 21% at SS of 0.07% and overestimated by less than 8% within the SS range of 0.1% to 0.7%. Additionally, the overestimation of Nc leads to an overestimation of rc, resulting in an underestimation of Tauto. As shown in Fig. 7d, using κ = 0.3 results in an underestimation of Tauto ranging from −19% to −56% within the SS range of 0.07% to 0.7%. Conversely, employing the κ(D) parameterization from Eq. 2 leads to an overestimation of Tauto by 21% at SS of 0.07% and an underestimation by less than 8% within the SS range of 0.1% to 0.7%. Hence, concerning clouds within the STP region, the employment of κ = 0.3 amplifies the Twomey effect and prolongs cloud lifetime through the attenuation of the collision-coalescence process. Conversely, when utilizing the κ(D) parameterization from Eq. 2, the predictive bias for AIEs is significantly reduced. It is worth emphasizing that, in contrast to high SS conditions, the deviations in predicting τc and Tauto are more prominent when κ = 0.3 is employed under low SS conditions (e.g., SS < 0.2%). This underscores a more substantial influence on shallow clouds and fogs with relatively low values of SS e.g.,63,64,65.

Discussions

As the “Asian Water Tower”, clouds in the Tibetan Plateau have a substantial impact on regional water resource distribution and the global climate37,38. As the important physical process of cloud formation, the characteristics of aerosol activation significantly influence the microphysical properties of clouds in the TP region. However, current research on aerosol activation in the TP region is nearly non-existent, with scarcely documented studies, contributing to deviations in cloud and precipitation simulations and climate prediction. As part of the Second Tibetan Plateau Scientific Expedition and Research Program, we conducted a comprehensive in-situ aerosol-cloud-precipitation experiment in the STP region. This study focuses on the observational results of aerosol activation characteristics, revealing the aerosol hygroscopicity and activation characteristics in the STP region. The specific results are summarized as follows.

The overall aerosol activation capacity in this region is weak, manifested by low NCCN and AF, along with weak hygroscopicity. In the SS ranges of 0.07% to 0.7%, the mean NCCN and AF spans from 24 to 483 cm−3 and 2 to 48%, respectively. Aerosol hygroscopicity, calculated through various measurements, including amalgamation of NCCN(SS) with dry PNSD, as well as the median particle size, volume concentration, and scattering coefficient of dry and wet aerosols, all indicate weak hygroscopicity with mean values below 0.1. This phenomenon may be attributed to the distinctive underlying surface and fuel usage practices among residents in the STP region55. Air masses in this region mainly originate from local emissions and the transport from its southern lower-altitude regions. With an increase in air mass transport distance, the proportion of air masses decreases, and the aerosol activation capacity weakens due to the effective CCN being depleted in the process of cloud and precipitation formation during the ascent along the terrain.

The κ(D) was derived through the integration of NCCN(SS) and dry PNSD, revealing a non-monotonic correlation between particle size and κ. As particle size increases, κ initially rises and subsequently decreases, reaching its peak at a diameter of 159.8 nm. The κ(D) relationship follows a Gaussian distribution, with fitting results suitable for κ parameterization to model aerosol activation. Utilizing this κ(D) parameterization induces a shift in the mean NCCN prediction bias from −51 to 26% as SS increases from 0.07% to 0.7%. Conversely, employing the recommended κ value of 0.3 for continental regions32 to predict cloud droplet activation in the TP region could result in an overestimation of NCCN and Nc by more than four times, especially at the SS of 0.1%. This could further lead to a 74% overestimation of cloud optical thickness and a 56% underestimation of the cloud-rain autoconversion threshold function, suggesting a substantial overestimation of the aerosol indirect effects. These discrepancies can be effectively mitigated by adopting the κ(D) parameterization proposed in this study.

These findings reveal the characteristics of aerosol activation in the STP region, contributing to a refined comprehension of cloud formation in the TP region and an improvement of regional cloud precipitation and global climate simulations. Although these are in-situ observational results from a single station in the STP region, it is notable that aerosol chemical composition measurements conducted at various sites in the STP region consistently indicate a high proportion of organic aerosol and black carbon39. This implies that weak hygroscopicity is likely a prevalent feature in the STP region. Nevertheless, advocating for more in-situ experiments with a focus on aerosol activation characteristics, including the measurements of aerosol hygroscopicity, CCN, and ice nuclei, in the TP region is advisable. Such endeavors would contribute to a more comprehensive understanding of aerosol activation capacity in the region, ultimately refining the accuracy of simulations related to plateau cloud-precipitation and climate. Furthermore, the κ(D) parameterization proposed in this study requires validation through a more extensive and prolonged collection of hygroscopicity measurements from TP region and remains to be applied into models to evaluate its predictive performance on aerosol activation.

Methods

Experiment and data

As part of the Second Tibetan Plateau Scientific Expedition and Research Program, the primary objective of GACPE-STP is to conduct observations of aerosols, clouds, and precipitation in the TP region. The visual representations of the experimental square cabin and the instruments employed in this experiment are provided in Supplementary Fig. 4 and Supplementary Table 1. Positioned approximately 10 kilometers from residential areas, the site is enveloped by forests and meadows. The aerosol optical properties at this location have been reported in Tian et al.66. The goals of GACPE-STP encompass two key aspects. Firstly, it aims to address the deficiency in microphysical observations of aerosols, clouds, and precipitation in the plateau region, while also seeking to elucidate the mechanisms through which aerosols activate and lead to cloud formation and subsequent precipitation. Secondly, it endeavors to quantify aerosol-cloud interactions in the plateau region and reduce the predictive uncertainties associated with the aerosol indirect effect in this specific area.

The focus of this study revolves around the characteristics of aerosol activation, which is largely unexplored in the TP region. Supplementary Figure 5 provides a schematic diagram illustrating the experimental setup used for measuring aerosol activation. Two scanning mobility particle sizers (SMPS, model 3938L50; TSI Inc.) were employed to assess PNSD under both dry and ambient relative humidity (RH) conditions. The measured diameter ranged from 18.8 to 685.4 nm, divided into 101 bins, and PNSD data was collected at 2-minute intervals. It is worth noting that we installed a water storage bottle before the wet SMPS inlet to prevent cloud droplets from blocking the inlet of wet SMPS. A CCN counter equipped with double cloud columns (CCNC, model 200; DMT Inc.) was utilized to determine the NCCN at varying SS conditions. A constant SS of 0.2% was maintained in column A. In column B, a total of nine distinct SS conditions were cycled, i.e., 0.07%, 0.1%, 0.15%, 0.2%, 0.3%, 0.4%, 0.5%, 0.6%, and 0.7%. Among these SS conditions, the measurement was set at SS = 0.07% for a duration of 15 min, while the others were each set at 5 min. Only CCN data with a temperature-stabilized state are used for analysis, comprising approximately 70% of the total CCN dataset of column B. To simultaneously measure the dry scattering coefficient (σsc,dry) and the scattering coefficient at a specific RH (σsc,wet), an improved humidified nephelometer system (PB-FRH100; BMET Inc.) was employed. The RH cycle set for the wet nephelometer ranges from 70 to 91%, with a 25-minute cycle period. Its operating principle, hardware configuration, and the uncertainty of observation data were described in Kuang et al.67 and Zhao et al.68.

The instruments were meticulously calibrated throughout the experiment. The CCNC was calibrated following the procedures as outlined in Rose et al.69. The SS calibration curves for column A and B in CCNC are in Supplementary Fig. 6. The PB-FRH100 was subjected to cleaning and calibration procedures. Data quality was initially assessed by scrutinizing the consistency of measurements acquired from diverse instruments, and the results are presented in Supplementary Fig. 7. The Na measured by the two SMPS exhibited a high level of concordance, with an average deviation of a mere 7%. Meanwhile, the NCCN flowing through the two cloud columns at SS of 0.2% within the CCNC demonstrated a robust correlation. However, on average, NCCN in column A was 24% lower than that in column B. Although this deviation is well within the expected range for the instrument, it causes a 35% difference in κ calculation. Moreover, σsc,dry and σsc,wet at 525 nm recorded by the two nephelometers in the PB-FRH100 exhibited a strong and consistent correlation. An increase in humidity had a minor effect on Na but a significant influence on aerosol volume concentration (Va) and σsc due to the hygroscopic growth of aerosols. Therefore, it is reasonable to observe higher σsc,wet compared to σsc,dry. Furthermore, significant correlations were observed between dry Na and NCCN in column A, as well as between dry Na and σsc,dry, affirming the expected relationship that increased Na correspondingly yields higher NCCN and enhanced scattering effects. As such, these results affirm the relatively high quality of the data.

Methods for calculating κ

Based on the observations provided by these instruments, we applied four distinct methods to calculate aerosol κ, each offering a unique perspective. The κ derived from these diverse methods can undergo cross-validation. The schematic diagram for the four methods is provided in Supplementary Fig. 8. The four methods can be categorized into two groups: the first involves the calculation of κ through the integration of NCCN(SS) and dry PNSD, while the second pertains to the determination of κ based on the concurrent measurements of both dry and wet aerosol properties. Details of the algorithms are as follows.

κ derived from N CCN (SS) and dry PNSD

Particle size matters more than chemistry for aerosol activation ability57 and the humidity requirement for activation decreases with increasing dry particle size14,16. Affected by aerosol mixing, the AF exhibits a gradual increase with dry particle diameter, following a sigmoidal pattern rather than an abrupt step function. The Dc is subsequently defined as the dry particle size at which 50% of the dry particles activate at a specified SS.

In the case of assuming internal mixing, Dc can be determined by integrating the dry PNSD from larger to smaller particle sizes until it matches the NCCN (SS) as

$${N}_{{CCN},{SS}}={\int }_{{\!D}_{c,{SS}}}^{{D}_{\max }}{PNSD}\,\left(D\right)\,{dD}$$
(8)

where D is dry particle diameter, Dmax represents the maximum D, which, in this context, is 685 nm as determined by the SMPS. Linear interpolation was employed in cases where Dc did not align with the lower or upper diameters of each bin in the PNSD.

Then hygroscopicity (κccn) can be determined by taking the derivative of the κ-Köhler equation16 with respect to Dc and SS:

$${\kappa }_{{ccn}}=\frac{4{A}^{3}}{27{{D}_{c}}^{3}{\mathrm{ln}}^{2}\left(1+{SS}/100\right)}$$
(9)

where A can be considered a function of the absolute temperature. In this study, we assumed σs/a to be that of pure water, specifically 0.0728 Nm−2. Ongoing debates surround the significance of σs/a variations and the associated bulk/surface partitioning in the activation of aerosols70 which is not focused on in this study. Various SS values were set within column B in the CCNC, yielding NCCN(SS). This NCCN(SS) data was subsequently utilized in Eq. 8 to derive Dc(SS), which was then employed in Eq. 9 to calculate the κccn at each Dc(SS).

This algorithm determines Dc and κccn from collocated polydisperse CCN and dry PNSD measurements, similar methodology that has been employed in prior studies e.g.,23,31,71. The limitations of this algorithm are primarily associated with the inherent bias in Na and NCCN measurements obtained from the SMPS and CCNC. To mitigate this uncertainty, one solution is to employ a monodisperse CCN measurement system27. Nevertheless, due to the low background Na at the experimental site, the Na for specific particle sizes, as isolated using the monodisperse differential mobility analyzer (DMA), and the corresponding NCCN often fall below 1 cm³, introducing substantial measurement bias. As such, the polydisperse CCN measurements were chosen for both columns A and B of CCNC within this experiment. Additionally, it’s worth noting that Eq. 9 is derived from an approximation of the κ-Köhler equation, and it may introduce a slight bias in the calculation of κccn when dealing with particles that are less-hygroscopic16.

Furthermore, CCN measurements may include unactivated particles that are incorrectly classified as CCN46,47, resulting in an overestimation of NCCN, especially at low SS conditions. Wang et al.46 introduced an approach based on the inverse relationship between the critical wet activation diameter of droplets (Dc,wet) and the critical SS to identify unactivated particles and correct the CCN measurements. Tao et al.72 further considered the impact of kinetic limitations and improved the determination of Dc,wet using a validated kinetic model73. In this study, we applied the Dc,wet from Tao et al.72 with assuming an accommodation coefficient of water vapor of 0.2 to rectify the NCCN measurements at low SS conditions.

κ derived from dry and wet aerosol properties

The impact of aerosol hygroscopic growth is noticeable in the increase of wet particle sizes, resulting in higher Va and σsc. The hygroscopic growth factor for particle diameter (g(RH)) is defined as the ratio of wet particle diameter at a specific RH to the corresponding dry diameter, expressed as:

$$g({RH})=\frac{D({RH})}{D({{RH}}_{{dry}})}$$
(10)

The κ-Köhler equation, expressed in terms of g, is given in Eq. 11 below:

$$\frac{{RH}}{100}=\frac{{g}^{3}-1}{{g}^{3}-(1-{\kappa }_{g})}\exp \left(\frac{A}{D({RH})}\right)$$
(11)

Given that RH at the dry condition (RHdry) is not zero in dry SMPS, κg can be derived as

$${\kappa }_{g}=({1-g}^{3})\Big/\left[\frac{{g}^{3}{{RH}}_{{dry}}}{100\exp \left(\frac{A}{D\left({{RH}}_{{dry}}\right)}\right)-{{RH}}_{{dry}}}-\frac{{RH}}{100\exp \left(\frac{A}{D\left({RH}\right)}\right)-{RH}}\right]$$
(12)

The derivation of Eq. 12 is provided in Supplementary Method. By utilizing the median diameters (Dmedian) of the dry and wet PNSDs measured by two SMPS, g(RH) and subsequently the related κ, κg, can be calculated.

The hygroscopic growth factor for light scattering (f(RH)) is defined as the ratio of σsc at a specific RH to the corresponding dry σsc, expressed as:

$$f({RH})=\frac{{\sigma }_{{\rm{s}}c}({RH})}{{\sigma }_{{\rm{s}}c}({{RH}}_{{dry}})}$$
(13)

In this study, σsc at 525 nm was used to calculate f(RH). Based on the κ-Köhler equation, Brock et al.74 introduced a single-parameter representation to characterize f(RH), and Kuang et al.67 extended this by considering that RHdry is not zero in dry nephelometer, leading to the following formulation:

$$f\left({RH}\right)=\left(1+{\kappa }_{{sc}}\frac{{RH}}{100-{RH}}\right)\Big/\left(1+{\kappa }_{{sc}}\frac{{{RH}}_{{dry}}}{100-{{RH}}_{{dry}}}\right)$$
(14)

where κsc is a parameter that best fits the f(RH) relationship. It is worth noting that the assumed linearly positive relationship between Va and σsc results in κsc not entirely conforming to the conventional interpretation of aerosol κ. To address this, Kuang et al.67 introduced a method for calculating the overall aerosol κ derived from measured f(RH), denoted as κf. κf is influenced by κsc and the Ångström exponent. The Ångström exponent can be determined using σsc values at 450 nm and 525 nm measured by the dry nephelometer in PB-FRH100. Subsequently, κf can be calculated. Note that the calculation of κsc in Eq. 14 does not account for the curvature effect, introducing slight bias in the subsequent calculating κf.

Additionally, the total Va can be calculated by assuming spherical particles. The hygroscopic growth factor for Va (V(RH)) is defined as the ratio of wet Va at a specific RH to the corresponding dry Va, expressed as:

$$V({RH})=\frac{{V}_{a}({RH})}{{V}_{a}({{RH}}_{{dry}})}$$
(15)

In a similar manner to Eq. 14, volume-weighted mean κ, κv, is calculated using V(RH) as

$$V\left({RH}\right)=\left(1+{\kappa }_{v}\frac{{RH}}{100-{RH}}\right)\Big/\left(1+{\kappa }_{v}\frac{{{RH}}_{{dry}}}{100-{{RH}}_{{dry}}}\right)$$
(16)

It is crucial to emphasize that the RH output from the SMPS pertains to the RH of the sheath gas that has passed through the heat exchanger, rather than the RH of the sample gas. In this study, we relied on the RH obtained from the automated weather station (WXT530, Vaisala Inc.) as the ambient RH, and the RH measured in the dry nephelometer of PB-FRH100 as the RHdry. It should be noted that there are deviations between the ambient RH and the RH of the wet sample gas. During clouds (fogs) formation, water accumulates in the storage bottle, resulting in both the ambient RH and the RH at the inlet of the wet SMPS being high, with little disparity between them. However, in the absence of clouds (fogs) and no water accumulation in the bottle, the ambient RH may be slightly higher than the wet SMPS inlet RH due to moisture loss in the sample gas pipeline. These deviations in RH can lead to deviations in the calculation results of κv and κg. As shown in Supplementary Fig. 9, a 10% increase or decrease in the absolute value of the sample gas RH leads to a respective halving or doubling of κv (and κg). Assuming a deviation of ±10% between the RH at the wet SMPS inlet and the ambient RH, the mean range for κv is 0.015 to 0.099, and for κg is 0.025 to 0.122, still indicating weak hygroscopicity. For future experiments, it is advisable to install a hygrometer between the water storage bottle and the wet SMPS inlet to precisely measure the RH of the wet sample gas, thereby mitigating biases in κv and κg calculations.

HYSPLIT model

Utilizing three-day backward trajectories generated by the HYSPLIT model75 (https://www.arl.noaa.gov/hysplit/, last access: 30 October 2023), we calculated a total of 1464 trajectories throughout the experimental period. The estimated lifetime of the measured aerosols is approximately 3 days. Consequently, the period for calculating backward trajectories was set to 3 days, a common practice in the majority of aerosol studies conducted in the TP region e.g.,76,77. These trajectories were initiated at a height of 100 m and calculated at hourly intervals. Subsequently, we identified three clusters from the 1464 trajectories, as illustrated in Fig. 1. The prevailing air masses predominantly originated from the southern regions with relatively lower altitudes. Roughly two-thirds of the air masses originated from local emissions and short-range transport within a 100-kilometer distance to the southwest (C1). Air masses from the southeast, associated with medium-range transport, constituted 31% of the total (C2), while air masses from the southwest, characterized by long-range transport, were the least common, accounting for 4% (C3). Abundant water vapor from the Indian Ocean, lifted by the terrain as the air mass travels northward, frequently results in forming clouds and precipitation at the observation site.