Measurement report: Size-resolved and seasonal variations in aerosol hygroscopicity dominated by organic formation and aging: insights from a year-long observation in Nanjing

Aerosol hygroscopicity plays a significant role in atmospheric chemistry, radiation, and climate effects. While previous studies have investigated regional differences in aerosol hygroscopicity, long-term observational studies focusing on seasonal variations in specific regions remain scarce. This study explores size-resolved and seasonal variations in aerosol hygroscopicity in northern Nanjing, using one-year hygroscopicity-tandem differential mobility analyser (H-TDMA) measurements in 2021. Aerosols in the region show relatively low hygroscopicity due to a high organic content (annual average mass fraction: 42.92 % in PM_2.5) in fine particles. The mean hygroscopicity parameter (κ_mean) increases with particle size across all seasons. Particles (40–200 nm) show seasonal κ_mean variations: winter (0.12–0.24) and spring (0.14–0.25) display relatively higher values attributable to relatively higher secondary inorganic content, while summer (0.12–0.21) and autumn (0.10–0.20) exhibit relatively weaker hygroscopicity due to enhanced contributions from less hygroscopic components. Diurnal patterns are shaped by photochemical aging and aqueous-phase reactions, leading to κ_mean slight enhancement for larger particles in the afternoon and evening. New particle formation (NPF) events occur most frequently in spring. During spring NPF days, Aitken-mode particles exhibit slightly low hygroscopicity, whereas accumulation-mode particles demonstrate relatively higher hygroscopicity compared to non-NPF days. Regional transport analysis reveals distinct controlling factors: hygroscopicity of 40 nm particles may be mainly controlled by local sources, while 200 nm particles are more influenced by seasonal air mass transport. These results improve understanding of aerosol–cloud interactions and support regional climate modeling and air quality management in urbanizing areas.

1 Introduction

Aerosols, defined as mixtures of solid and liquid particles suspended in the air, possess hygroscopicity as a fundamental physicochemical property that governs their interactions with water vapor under varying environmental relative humidity (RH). The hygroscopicity of aerosols describes their capacity to absorb water vapor, significantly influencing both the atmospheric environment and global climate through complex physicochemical processes (Chen et al., 2019; Zhang et al., 2023).

Aerosol hygroscopicity plays a central role in determining the phase state, chemical reactions, optical properties, and cloud nucleation activity of aerosol particles, thereby significantly influencing atmospheric chemistry, radiation, and climate effects (Chen et al., 2022; Peng et al., 2020; Ray et al., 2023; Swietlicki et al., 1999; Y. Wang et al., 2025; Zhang et al., 2023). Hygroscopic growth notably alters aerosol optical properties by modifying particle size distribution and refractive index, enhancing light scattering coefficient (Liu et al., 2022; Song et al., 2023). Furthermore, hygroscopicity promotes cloud formation by activating particles as cloud condensation nuclei, which leads to substantial changes in radiative forcing patterns (Rosenfeld et al., 2014; Svenningsson et al., 2006). However, aerosol hygroscopicity is influenced by various factors. Variations in environmental conditions and physicochemical processes can result in divergent hygroscopic behaviors across different atmospheric environments (Gysel et al., 2007; Jiang et al., 2025; N. Wang et al., 2025).

Currently, multiple instruments and techniques are available for measuring aerosol hygroscopicity. Among the more established instruments are the hygroscopicity-tandem differential mobility analyser (H-TDMA), the cloud condensation nuclei counter (CCN_C), and the dual-nephelometer system (Chen et al., 2023; Jin et al., 2022; Song et al., 2023; Wang et al., 2017; Zhang et al., 2017). Compared to other hygroscopicity measurement instruments, the H-TDMA, which is based on measuring particle number concentration, offers distinct advantages in studying the properties of ultrafine mode particles (Wang et al., 2019). In addition to measuring the size-resolved hygroscopic growth factor (GF) of aerosols under varying relative humidity conditions, the H-TDMA can also provide insights into the mixing state of particles and reflect aerosol chemical compositions (Chen et al., 2022; X. Wang et al., 2018). According to Köhler theory, GF depends on chemical composition (Raoult effect) and particle size (Kelvin effect) (Petters and Kreidenweis, 2007). Compared with GF, hygroscopicity parameter (κ) introduced by Petters and Kreidenweis (2007) eliminates the influence of the Kelvin effect, enabling direct comparison of hygroscopicity governed by the Raoult effect across different particle sizes (Kammermann et al., 2010; Petters and Kreidenweis, 2007).

A more profound understanding of aerosol hygroscopicity is crucial for improving the predictive capability of global climate models, particularly in simulating aerosol size distributions and their scattering properties under varying humidity conditions. The size dependence of aerosol hygroscopicity exhibits significant variations across different atmospheric environments (Peng et al., 2020; Zhang et al., 2023). To better characterize the spatiotemporal evolution of aerosol hygroscopicity under diverse pollution conditions, extensive and comprehensive studies on its spatiotemporal variability are required. Furthermore, understanding the effects of atmospheric processes, including new particle formation (NPF) and haze events, on aerosol hygroscopicity across heterogeneous environmental regimes remains imperative (Chen et al., 2022).

NPF, defined as the evolution process where newly formed sub–3 nm particles grow to larger sizes, represents a significant atmospheric aerosol source capable of influencing aerosol hygroscopicity and has been observed globally (Hirshorn et al., 2022; Shen et al., 2023; Yli-Juuti et al., 2011). Generally, freshly nucleated particles exhibit lower hygroscopicity, and the hygroscopicity increases with particle aging (Asmi et al., 2010). NPF events dominated by different chemical components exert distinct impacts on aerosol hygroscopicity. Newly formed sulfate-dominated particles exhibit higher hygroscopicity, whereas organics-dominated counterparts display lower hygroscopicity (Hong et al., 2024; Ma et al., 2016). Furthermore, Liu et al. (2021) observed that on NPF events, photochemical oxidation triggered by the nucleation of volatile organic compounds generates more water-soluble organic acids, resulting in higher hygroscopicity of organic aerosols on NPF days compared to non-NPF days.

In recent years, numerous studies on aerosol hygroscopicity based on H-TDMA observation data have been conducted globally. It has been found that the probability density function of GF or κ (GF/κ-PDF) exhibits a bimodal distribution at urban sites (Shi et al., 2022; Spitieri et al., 2023; Tan et al., 2013; X. Wang et al., 2018), while at some non-urban sites, it shows a unimodal or quasi-unimodal distribution (Chen et al., 2022; Y. Wang et al., 2018). Wang et al. (2019) discovered that ultrafine-mode particles during clean periods mainly originate from nucleation events followed by growth, whereas during severe pollution periods, they predominantly come from primary emissions in urban environments. Conversely, accumulation-mode particles are primarily from primary emissions during clean periods and secondary processes during pollution periods, resulting in notable differences of aerosol hygroscopicity for different mode particles under varying pollution levels. Additionally, aerosol hygroscopicity exhibits substantial differences among air masses of varying origins. Over the Antarctic continent, dry continental air masses are reported to exhibit stronger hygroscopicity than moist marine air masses (Asmi et al., 2010). At an urban site in Beijing, seasonal hygroscopicity variations are strongly correlated with air mass source regions (Zhang et al., 2023).

Aerosol concentration and composition undergo significant variations on both temporal and spatial scales. Long-term measurements of aerosol hygroscopicity are crucial for understanding its seasonal and annual variations, as well as its impacts on visibility, atmospheric chemistry, and climate change (Peng et al., 2020). In some regions of the world, studies utilizing H-TDMA for long-term measurements of aerosol hygroscopicity have already been conducted (Alonso-Blanco et al., 2019; Fors et al., 2011; Kammermann et al., 2010; Mamali et al., 2018; Ray et al., 2023). These observations reveal that aerosol hygroscopicity is highly depending on their sources and physicochemical aging processes. In China, numerous observational experiments measuring aerosol hygroscopicity using H-TDMA have been carried out in regions such as the North China Plain (NCP), the Yangtze River Delta (YRD), and the Pearl River Delta (PRD) (Jiang et al., 2016; Jiang et al., 2025; Wang et al., 2017). However, most measurements limit to short-term field campaigns (typically 1–2 months duration), which makes it impossible to determine the seasonal variations and their influencing factors in aerosol hygroscopicity at specific locations (Fan et al., 2020).

In this study, the H-TDMA system is utilized in the northern suburbs of Nanjing to obtain size-resolved hygroscopicity observation data for submicron aerosols, covering the entire year from January to December 2021. The H-TDMA observations enable the determination of size-resolved and seasonal variations in aerosol hygroscopicity in the Nanjing region. Furthermore, combining these data with aerosol chemical composition measurements facilitate the further analysis of influencing factors contributing to these hygroscopicity differences. This paper is structured as follows. Section 2 describes the instrumentation and the methods to data analysis. Aerosol hygroscopicity during different seasons are discussed in Sect. 3. Conclusions and summary are given in Sect. 4.

2 Experiment and data analysis

2.1 Measurement site and campaign

A comprehensive field observation experiment was conducted in the northern suburbs of Nanjing in 2021, aiming to delve deeply into the interactions among the atmosphere, aerosol, boundary layer, and cloud. The observation site is located on the campus of Nanjing University of Information Science and Technology (NUIST, 32°13^′ N, 118°46^′ E) in the northern suburban area of Nanjing, which is situated in the central YRD. Further details on the field campaign and measurement site are available in previous work (Song et al., 2023; Wang et al., 2016; Yang et al., 2025).

This work provides a comprehensive examination of seasonal variations in aerosol hygroscopicity and their relationship with chemical composition, based on year-round observational data collected from January to December 2021. The dataset is partitioned according to conventional meteorological seasons: winter (January–February and December 2021), spring (March–May 2021), summer (June–August 2021), and autumn (September–November 2021).

2.2 Measurements and data analysis

2.2.1 Measurement of aerosol hygroscopicity

The H-TDMA used in this study primarily consists of two differential mobility analysers (DMA, Model 3081L, TSI Inc.) and a condensation particle counter (CPC, Model 3772, TSI Inc.). The dried and neutralized aerosol sample is first passed through the first DMA, which selects monodisperse particles of specific diameters (40, 80, 110, 150, and 200 nm). Subsequently, the sample with monodisperse particles is humidified to RH =90 % via a nafion humidifier. Finally, the sample is directed through the second DMA and the CPC to measure the particle number size distribution of the humidified particles.

The hygroscopic growth factor (GF) is defined as:

$$\begin{array}{}\text{(1)}& \mathrm{GF}={\displaystyle \frac{D_{\mathrm{p}}\left(\mathrm{RH}\right)}{D_{\mathrm{p}}\left(\mathrm{dry}\right)}},\end{array}$$

where D_p(dry) denotes the dry diameter of monodisperse particles selected by the first DMA before humidification, and D_p(RH) represents the particle diameter selected by the second DMA after humidification at RH =90 %.

The H-TDMA data can be used to compute the measured distribution function of GF for any selected D_p(dry) particles, which is then used to retrieve the probability distribution function of GF (GF-PDF) according to the multi-mode TDMAfit algorithm.

The hygroscopicity parameter (κ) is calculated as follows (Petters and Kreidenweis, 2007):

$$\begin{array}{}\text{(2)}& \mathit{\kappa }=({\mathrm{GF}}^{\mathrm{3}}-\mathrm{1})\cdot \left[{\displaystyle \frac{\mathrm{1}}{\mathrm{RH}}}\exp \left({\displaystyle \frac{\mathrm{4}{\mathit{\sigma }}_{\mathrm{s}/\mathrm{a}}{M}_{\mathrm{w}}}{RT{\mathit{\rho }}_{\mathrm{w}}{D}_{\mathrm{p}}\mathrm{GF}}}\right)-\mathrm{1}\right],\end{array}$$

where ${\mathit{\sigma }}_{\mathrm{s}/\mathrm{a}}$ is the surface tension of the droplet–air interface at the composition of the droplet, M_w is the molar mass of water, R is the universal gas constant, T is the temperature, and ρ_w is the density of water.

The probability distribution function of κ (κ-PDF, c (κ, D_p)) derived from GF-PDF is normalized by $\int c(\mathit{\kappa },D_{\mathrm{p}})\mathrm{d}\mathit{\kappa }=\mathrm{1}$. Based on the κ values, aerosol particles are categorized into three hygroscopic groups: nearly hydrophobic (NH, κ<0.1), less hygroscopic (LH, $\mathrm{0.1}\le \mathit{\kappa }<\mathrm{0}.$2), and more hygroscopic (MH, κ≥0.2). The MH groups mainly consist of inorganic species such as sulfates, nitrates, and ammonium salts. In contrast, the NH and LH groups are primarily composed of black carbon, insoluble organics, and partially soluble organics (Liu et al., 2011; Müller et al., 2017; Swietlicki et al., 2008).

The mean κ (κ_mean) is then defined as the number-weighted mean of κ-PDF over the κ range [a, b]:

$$\begin{array}{}\text{(3)}& {\mathit{\kappa }}_{\mathrm{mean}}=\underset{a}{\overset{b}{\int }}\mathit{\kappa }c\left(\mathit{\kappa },D_{\mathrm{p}}\right)\mathrm{d}\mathit{\kappa },\end{array}$$

where a and b represent the lower and upper integration limits of κ. For the ensemble κ_mean, the integral spans the entire domain, i.e., a=0 and b→∞. When computing κ_mean of NH, LH, and MH hygroscopic groups (κ_NH, κ_LH, and κ_MH), a and b are set according to the predefined κ range of each group.

Accordingly, the number fraction (NF) for each hygroscopic group with over the range [a, b] is defined as:

$$\begin{array}{}\text{(4)}& \mathrm{NF}=\underset{a}{\overset{b}{\int }}c\left(\mathit{\kappa },D_{\mathrm{p}}\right)\mathrm{d}\mathit{\kappa }.\end{array}$$

The standard deviation of κ-PDF (σ_κ-PDF) is calculated as:

$$\begin{array}{}\text{(5)}& {\mathit{\sigma }}_{\mathit{\kappa }\text{-PDF}}={\left(\underset{\mathrm{0}}{\overset{\mathrm{\infty }}{\int }}{\left(\mathit{\kappa }-{\mathit{\kappa }}_{\mathrm{mean}}\right)}^{\mathrm{2}}c\left(\mathit{\kappa },D_{\mathrm{p}}\right)\mathrm{d}\mathit{\kappa }\right)}^{\frac{\mathrm{1}}{\mathrm{2}}}.\end{array}$$

Previous research has widely adopted σ_κ-PDF as a metric for aerosol mixing state characterization (Jiang et al., 2016; Wang et al., 2017; Yuan et al., 2020). However, this absolute dispersion parameter fails to account for scenarios where significant κ_mean variations exist between different sizes. To address this limitation, this study introduces the coefficient of variation (CV_κ-PDF):

$$\begin{array}{}\text{(6)}& {\mathrm{CV}}_{\mathit{\kappa }\text{-PDF}}={\displaystyle \frac{{\mathit{\sigma }}_{\mathit{\kappa }\text{-PDF}}}{{\mathit{\kappa }}_{\mathrm{mean}}}}.\end{array}$$

This parameter effectively captures the relative dispersion of size-resolved κ-PDF by incorporating κ_mean differences among particle populations. Consequently, CV_κ-PDF serves as the primary mixing state indicator in this study, with lower values corresponding to stronger internal mixing.

2.2.2 Measurements of other aerosol properties

The measurement of aerosol chemical compositions is conducted using an aerosol chemical speciation monitor (ACSM, Aerodyne Research Inc.) coupled with an aethalometer (AE-33, Magee Scientific Inc.). Both instruments are configured with PM_2.5 sampling inlets: the ACSM system incorporates a PM_2.5 aerodynamic lens, while the AE-33 utilizes an aerosol cutter with 2.5 µm cutoff diameter. This configuration ensures measurement of particulate matter mass concentrations specifically for the aerodynamic diameter fraction below 2.5 µm (PM_2.5).

The ACSM operates at a temporal resolution of 15 min, while the AE-33 provides measurements at a higher resolution of 1 min. The ACSM, equipped with a capture vaporizer and a quadrupole mass spectrometer, is primarily used to measure the mass concentrations of organic aerosols (OA), sulfate (SO${}_{\mathrm{4}}^{\mathrm{2}-}$), nitrate (NO${}_{\mathrm{3}}^{-}$), ammonium (NH${}_{\mathrm{4}}^{+}$), and chloride (Chl). To further analyse OA composition, the positive matrix factorization analysis is applied to differentiate between primary organic aerosols (POA) and secondary organic aerosols (SOA). Concurrently, the AE-33 is employed to measure the mass concentration of black carbon (BC). The total PM_2.5 mass concentration is derived by summing the mass concentrations of all components measured by the ACSM and AE-33. The absence of dust-related measurements may lead to underestimation of PM_2.5 mass concentrations during dust episodes in this study.

The particle number size distribution (PNSD) is measured by two scanning mobility particle sizers (SMPS) covering different size ranges: Nano-SMPS (2–60 nm) and SMPS (15–700 nm), both operating with a temporal resolution of 5 minutes. The PNSD-related data presented in this study are all provided by SMPS. NPF events are identified using SMPS data based on a significant particle number increase around 20 nm followed by banana-shaped PNSD growth, with nano-SMPS data being used auxiliarily for confirmation despite frequent gaps. Particles are classified into two modes based on their diameter: ultrafine mode (D_p≤100 nm) and accumulation mode (100 nm $<D_{\mathrm{p}}\le \mathrm{700}$ nm). Specifically, particles within the 40 nm $<D_{\mathrm{p}}\le \mathrm{100}$ nm range are classified as belonging to the Aitken mode. Then the total number concentrations in ultrafine- and accumulation-mode particles (N_ult and N_acc) are then calculated, separately.

2.3 Backward trajectory calculation and clustering analysis

Backward trajectories of air masses arriving at the sampling site are calculated using the NOAA HYSPLIT-4 (Hybrid Single-Particle Lagrangian Integrated Trajectory) model (Draxler and Hess, 1998; Wu et al., 2016). The 72 h backward trajectories terminated at 100 m above ground level at 3 h intervals from 00:00 to 21:00 LT (Local Time, UTC+08). The number of clusters is determined based on the variation in total spatial variance (refer to the HYSPLIT4 User Guide), with the k-means clustering algorithm applied to classify trajectories for each season. To balance optimal trajectory separation (favoring a larger number of clusters) with visualization simplicity (preferring fewer clusters), air mass backward trajectories are partitioned into 3 clusters per season.

Figure 1 (a) Size variations of the mean hygroscopicity parameter (κ_mean) during different seasons, with error bands indicating the standard deviations of κ_mean. (b) Seasonal distributions of mass concentrations and compositional fractions for PM_2.5 chemical species.

3 Results and discussion

3.1 Overview

3.1.1 Seasonal variations of size-resolved aerosol hygroscopicity

As shown in Fig. 1a, a significant increase in κ_mean with particle size is observed in all seasons, consistent with a previous study in Nanjing (Jiang et al., 2025). The size-resolved annual κ_mean values are 0.12 ± 0.04, 0.16 ± 0.05, 0.18 ± 0.05, 0.20 ± 0.05, and 0.22 ± 0.06 for particles with diameters of 40, 80, 110, 150, and 200 nm, respectively (Table S1 in the Supplement). Similar hygroscopicity-particle size dependence is observed in Madrid, while more complex relationships are found in Athens, India and Switzerland (Alonso-Blanco et al., 2019; Kammermann et al., 2010; Ray et al., 2023; Spitieri et al., 2023). The enhancement of aerosol hygroscopicity with particle size may be attributed to chemical aging processes of particles and an increased proportion of inorganic components (Alonso-Blanco et al., 2019). Compared to the measurements in the NCP (Chen et al., 2022), κ at this site is relatively lower, likely due to the dominance of organic aerosols (annual average mass fraction: 42.92 % in PM_2.5) (Fig. 1b). This pattern closely resembles observations from Shanghai, which is also located within the YRD region (Chen et al., 2022).

A pronounced seasonal contrast in κ_mean is observed for 40 nm particles, with spring exhibiting the highest value (κ_mean=0.14) compared to other seasons (κ_mean=0.10–0.12). For larger particles (80–200 nm), obviously higher κ_mean values are recorded in winter and spring than in summer and autumn. This pattern may be attributed to the increased mass fractions of hydrophilic sulfate–nitrate–ammonium (SNA) during the more severe PM_2.5 pollution in winter and spring. As shown in Fig. 1b, the mass concentrations of PM_2.5 during winter and spring exceed 30 µg m⁻³, with corresponding relatively higher MF_SNA (the mass concentration of SNA) values of 49.37 % and 53.25 %, respectively. In contrast, summer and autumn measurements show relatively lower PM_2.5 levels, averaging approximately 25 µg m⁻³, with corresponding relatively lower MF_SNA values of 45.34 % and 47.12 %, respectively. Similar seasonal variation patterns in aerosol chemical composition are reported at another site in Nanjing during 2016–2017 (Xie et al., 2022).

In summary, aerosols in the northern suburban area of Nanjing exhibit relatively low hygroscopicity, which may be mainly due to their relatively higher organic content. A consistent increase in hygroscopicity with particle size is observed across all seasons. Furthermore, aerosols in winter and spring demonstrate enhanced hygroscopicity in the 80–200 nm size range compared to summer and autumn, which can be attributed to relatively higher content of SNA.

Figure 2 (a) Seasonal variations of NF_MH, NF_LH, and NF_NH (represented by orange, blue, and gray segments in the pie charts, respectively), alongside the corresponding κ_MH, κ_LH, and κ_NH (represented by orange, blue, and gray lines, respectively) for different size particles (40–200 nm). Error bars indicate the standard deviation of each parameter. (b) Seasonal variations of CV_κ-PDF for different size particles. Boxplots display the mean (red dots), median (horizontal line), 25th and 75th percentiles (lower and upper box), and 5th and 95th percentiles (lower and upper whiskers).

3.1.2 Seasonal variations of hygroscopic groups in the different size particles

Figure 2a illustrates the seasonal variations in the NF and κ of different hygroscopic groups (NH, LH, MH) across various particle sizes. With the exception of 40 nm particles, the size variations in κ_NH and κ_LH are less pronounced than those of κ_MH. Meanwhile, compared with NF_NH and NF_LH, NF_MH shows a relatively obvious increase with particle size. Considering the increase of κ_mean with particle size (Fig. 1a), this phenomenon suggests that the size variation in κ_mean may be mainly driven by MH group particles. Unlike particles in other size ranges, the 40 nm particles have relatively small seasonal differences in κ_NH, κ_LH, and κ_MH. The κ_mean of 40 nm particles in spring is relatively higher than in other seasons due to the lower NF_NH and higher NF_MH of 40 nm particles.

As shown in Fig. 2a, for Aitken-mode particles, κ_NH is found to increase consistently from winter and spring to summer and autumn, whereas for accumulation-mode particles, the lowest κ_NH values are recorded during summer. However, the variation in κ_NH remains limited, generally within the range of 0.03–0.05. Furthermore, κ_LH for 40–150 nm particles is slightly higher in spring compared to other seasons, while for 200 nm particles, it reaches slightly higher values in summer. Despite these variations, the variation in κ_LH remains minor, typically ranging between 0.14–0.17, which is in the κ range of SOA (Petters and Kreidenweis, 2007). Across all particle sizes, both κ_MH and NF_MH are bigger in winter and spring than in summer and autumn for the same particle sizes. Considering the higher κ_mean in winter and spring compared to summer and autumn (Fig. 1a), this suggests that the seasonal in κ_mean is likely driven by MH group particles.

In general, aerosol particles exhibit higher κ_MH in winter than in spring, which may be attributed to the higher abundance of NO${}_{\mathrm{3}}^{-}$ in winter aerosols leading to enhanced hygroscopicity of MH groups. This effect is more pronounced in accumulation-mode particles compared to Aitken-mode particles, with κ_MH showing more distinct seasonal variations in the former. However, a relatively higher NF_MH is observed in spring compared to winter due to the greater abundance of SNA, leading to a relatively higher κ_mean in spring (Fig. 1). Although the differences in aerosol κ_MH between summer and autumn are relatively small, the higher NF_MH in summer results in greater κ_mean across all particle size ranges compared to autumn.

Figure 3 The correlation coefficients (R²) between κ_mean and six parameters (NF_MH, NF_LH, NF_NH, κ_MH, κ_LH, κ_NH) for different size particles (40–200 nm) during different seasons.

As demonstrated in Fig. 2b, the CV_κ-PDF exhibits a decreasing trend with increasing particle size within the same season, reflecting enhanced internal mixing during aerosol aging from the Aitken mode to the accumulation mode. Notably, the CV_κ-PDF values of particles in winter are relatively higher compared to other seasons, and this phenomenon is particularly evident for 40 nm particles. This indicates that, relative to other seasons, particles in winter exhibit a relatively pronounced external mixing state. This seasonal pattern likely results from reduced photochemical activity during winter months, which inhibits atmospheric aging processes and promotes the persistence of externally mixed aerosols.

To further elucidate the impact of different hygroscopic groups on κ_mean, the correlation coefficients (R²) between the κ_mean and six parameters (NF_MH, NF_LH, NF_NH, κ_MH, κ_LH, and κ_NH) are calculated (Fig. 3). The results indicate that κ_mean is predominantly influenced by NF_MH in all seasons, with the corresponding R² values all above 0.69. For Aitken-mode particles, κ_mean is also affected largely by NF_NH, particularly for 40 nm particles. In autumn, the R² between NF_NH and κ_mean (R²=0.77) for 40 nm particles is even slightly higher than that between NF_MH and κ_mean (R²=0.71). This phenomenon may be attributed to the fact that Aitken-mode particles, primarily originating from direct emissions or NPF events with limited aging, are typically composed of more hydrophobic matters (e.g., OA and BC) (Gysel et al., 2007; Li et al., 2023). These particles are characterized by higher NF_NH and lower NF_MH, leading in an amplified influence of NF_NH on the κ_mean of Aitken-mode particles.

Overall, these findings suggest that variations in the NF of different hygroscopic groups have a greater impact on κ_mean than variations in the κ values of these groups. Beyond the influence by the NF of different hygroscopic groups, κ_mean for relatively larger particles is also affected by κ_MH, while for relatively smaller particles, due to their relatively weaker hygroscopicity, κ_mean is jointly influenced by κ_MH and κ_LH. Compared to winter and spring, the influence of κ_MH on κ_mean is smaller in summer and autumn, whereas the influence of κ_LH on κ_mean is greater. This pattern is attributed to the κ_mean for the same particle sizes is lower in summer and autumn than in winter and spring.

Figure 4 Diurnal variations of (a1–a4) particle number size distributions (PNSD), total particle number concentration (N_tot), accumulation-mode particle number concentration (N_acc), and ultrafine-mode particle number concentration (N_ult), (b1–b4) mass fractions (MF) of aerosol chemical species, and (c1–c4) mass concentrations (M) of aerosol chemical species during different seasons.

3.1.3 Seasonal variation of diurnal cycles in aerosol hygroscopicity

As illustrated in Fig. 4a1–a4, rush-hour-induced enhancements of particle number concentration demonstrate distinct seasonal variability, with pronounced effects on ultrafine-mode particles during morning (∼ 06:00 LT) and evening (∼ 20:00 LT) periods. This phenomenon is most pronounced in winter and spring, which may be related to the relatively stagnant meteorological conditions during these seasons. Quantitative analysis reveals relatively obvious PM_2.5 mass accumulation during rush hours across all non-summer seasons, accompanied by synchronous increases in both BC mass concentrations and fractions (M_BC and MF_BC) (Fig. 4). Relative to ultrafine-mode particles, the number concentration of accumulation-mode particles has a greater impact on PM_2.5 mass concentration. In contrast to other seasons, summer noon shows a more noticeable increasing trend in the number concentration of accumulation-mode particles, which to some extent weakens the diurnal variation amplitude of PM_2.5 mass concentration in summer, resulting in less pronounced changes in PM_2.5 mass concentration during summer rush hours (Fig. 4c3). Traffic-related compositional changes during rush hours, especially increased NF_NH, systematically reduce κ_mean across most particle size ranges, with magnitudes varying by particle size and season (Fig. 5). Compared to the Nanjing site in this study, sites like Madrid and Budapest show a more pronounced decline in κ_mean during traffic emissions rush hours (Alonso-Blanco et al., 2019; Enroth et al., 2018). This disparity implies that aerosol hygroscopicity at the Nanjing site is less sensitive to rush-hour traffic emissions. Notably, while PM_2.5 mass concentration increases are less significant during summer rush hours than in other seasons (Fig. 4c3), the reductions in κ values across most particle size ranges are more pronounced. This observation likely stems from the relatively low total particle number concentration (N_tot) in summer, whereby traffic emissions during morning and evening rush hours represent a larger proportion of N_tot, thereby inducing a relatively pronounced decrease in κ_mean.

Figure 5 Diurnal variations of the NF_MH (orange bars), NF_LH (blue bars), and NF_NH (gray bars), and κ_mean (black dots) for different size particles (40–200 nm) during different seasons.

Figure 4b1–b4 reveals daytime increasing trend in the mass fractions of SO${}_{\mathrm{4}}^{\mathrm{2}-}$ and SOA mass fractions across all seasons, with peak values coinciding with periods of maximum photochemical activity – a trend most pronounced in summer due to enhanced solar radiation (Peng et al., 2017). In contrast, the mass fraction of NO${}_{\mathrm{3}}^{-}$ displays an inverse diurnal cycle during non-winter periods, reaching maximal mass fractions during nocturnal periods. This behavior may stem from lower nocturnal temperatures and higher RH, which promote efficient aqueous-phase reactions and gas-particle partitioning, potentially involving NO₃ radical chemistry, converting gaseous precursors like NO_x into particulate NO${}_{\mathrm{3}}^{-}$ (Sun et al., 2013).

These compositional shifts drive distinct hygroscopicity dynamics primarily in accumulation-mode particles (e.g., 200 nm), as their composition aligns more closely with bulk PM_2.5 measurements (Wang et al., 2019). As shown in Fig. 5, during non-winter periods, variations in NF among hygroscopicity groups result in two chemically-distinct diurnal patterns for κ_mean in 200 nm particles: an afternoon enhancement (14:00–18:00 LT), likely driven by photochemical production of hydrophilic SO${}_{\mathrm{4}}^{\mathrm{2}-}$ and SOA, and a nighttime enhancement (post-20:00 LT), associated with NO${}_{\mathrm{3}}^{-}$ accumulation under favorable nocturnal conditions. This pattern is less evident in winter, where stagnant meteorological conditions suppress pollutant dispersion, leading to consistently high NO${}_{\mathrm{3}}^{-}$ levels and minimal diurnal variation in NO${}_{\mathrm{3}}^{-}$ mass fraction (Fig. 4).

3.2 Impact of NPF events on the size-resolved aerosol hygroscopicity

Figure 4a2 shows that the spring afternoon period exhibits unique particle dynamics, with the size distribution displaying a subtle banana-shaped diurnal pattern (more evident in event-specific examples like Fig. S1a1 in the Supplement). This pattern is potentially linked to the frequent NPF events that occur in spring. The frequency of NPF occurrences reaching 21 % during spring (approximately double the annual average), significantly higher than in other seasons (Table S2). Previous studies have shown that strong winds during prevalent spring dust episodes in China significantly scavenge fine particles, creating a cleaner environment conducive to NPF occurrence (Shen et al., 2023; Xu et al., 2022). In contrast to spring non-NPF days, the spring NPF days shows more pronounced diurnal fluctuations in the number concentration of ultrafine-mode particles and the mass concentration of PM_2.5 chemical compositions (Fig. S1). As shown in Fig. S1c1, around 10:00 LT in the spring NPF days, the mass concentration of PM_2.5 sharply decreases. Under these low pollution conditions, the lower condensation sink (CS) facilitates the onset of NPF events (Hong et al., 2023). This leads to a rapid increase in N_ult (Fig. S1a1).

In comparison to spring non-NPF days, MF_OA significantly increases in spring NPF days, with MF_SOA gradually increases over time. This trend starts at 10:00 LT and continues until 18:00 LT to the evening rush hours (Fig. S1b1). This suggests that spring NPF events may be predominantly driven by the formation of SOA. Additionally, in contrast to spring non-NPF days, a slightly lower κ_mean is observed for 40 nm particles (Fig. S2a1), suggesting that the Aitken-mode particles originating from NPF events exhibit lower hygroscopicity. This is related to a higher NF_NH and a lower NF_MH (Fig. S2b1 and d1). As Aitken-mode particles grow during spring NPF days, the κ_mean of 40 nm particles shows a modest increase from 08:00 to 12:00 LT, while the κ_mean increase of 80 nm particles is relatively lagged from 10:00 to 14:00 LT (Fig. S2a1), though differences in κ_mean for 40 nm particles between NPF and non-NPF days are minimal due to their generally low and less variable values.

Between 14:00–16:00 LT, the N_ult reaches its peak, while the N_acc and the PM_2.5 mass concentration reach their diurnal lows (Fig. S1). A slightly increasing trend in κ_mean occurs around 16:00 LT for most size particles, driven by concurrent increases in both the NF_MH and κ_MH (Figs. S2 and S3). This enhancement is particularly evident for accumulation-mode particles, which have a higher κ_MH. These findings suggest that Aitken-mode particles originating from NPF events, which exhibit lower hygroscopicity, gradually age and grow into more hygroscopic accumulation-mode particles over time during spring NPF days.

Figure 6 (a) κ_mean, κ_MH, κ_LH and κ_NH in spring NPF days and spring non-NPF days, and the κ_mean in winter for all size particles (40–200 nm). (b) NF_MH, NF_LH, NF_NH in spring NPF days and spring non-NPF days for all size particles (40–200 nm).

As shown in Fig. 6a, the κ_NH, κ_LH and κ_MH exhibit minimal difference between NPF days and non-NPF days during spring. However, the κ_mean on NPF days is significantly lower by 6.76 % for 40 nm particles and higher by 5.07 % for 200 nm particles compared to non-NPF days (Table S3). This divergence is attributed to the changes in the NF_NH and NF_MH: NPF days have larger NF_NH and smaller NF_MH for 40 nm particles, while the opposite is observed for 200 nm particles (Fig. 6b). Additionally, the diurnal variation of CV_κ-PDF for 40 nm particles in NPF days has relatively obvious fluctuation, while the CV_κ-PDF for other particle sizes in NPF days and all particle sizes in non-NPF days has no obvious diurnal variation (Fig. S3a1–a4). This indicates that the NPF event may only have a relatively pronounced effect on the internal mixing state of 40 nm particles. Furthermore, the CV_κ-PDF for Aitken-mode particles on NPF days is slightly larger than on non-NPF days, while the opposite is observed for accumulation-mode particles (Table S3). These findings suggest that, compared to non-NPF days, the internal mixing of relatively smaller particles is slightly weakened on NPF days, while relatively larger particles exhibit a slightly enhanced state of internal mixing.

Similar to spring, winter exhibits the relatively higher κ_mean values for particles in the 40–200 nm size range compared to summer and autumn (Fig. 1a). To further investigate seasonal differences, the κ_mean values of spring NPF days and non-NPF days are compared with winter conditions (Fig. 6a). For Aitken-mode particles, winter consistently shows lower κ_mean values than both spring NPF and non-NPF days. For accumulation-mode particles (except for 110 nm), winter κ_mean values are significantly lower than spring NPF days but comparable to spring non-NPF days, with only a slight lower in κ_mean relative to spring non-NPF days. Particles with a diameter of 110 nm represent a transitional size between the Aitken and accumulation modes. Their hygroscopicity in winter is slightly higher than that observed during spring NPF days, but slightly lower than during spring non-NPF days. These findings collectively demonstrate that, compared to spring non-NPF days, spring NPF days exhibit slightly reduced hygroscopicity in Aitken-mode particles and slightly enhanced hygroscopicity in accumulation-mode particles. Regardless of the influence of NPF events, particles of all sizes in spring demonstrate relatively stronger hygroscopicity than those of the same size in other seasons, except for 110 nm particles, which show slightly lower hygroscopicity during spring NPF days compared to winter.

Figure 7 72 h air mass backward trajectories at a height of 100 m corresponding to the cluster analysis during different seasons. The line colors denote different clusters, i.e., red for Cluster C1, green for Cluster C2, and blue for Cluster C3.

3.3 Impact of regional transport on aerosol hygroscopicity

3.3.1 Seasonal impacts of regional transport on aerosol hygroscopicity

Under the influence of the subtropical monsoon, seasonal variations in air mass sources are evident: while summer air masses derive from southern areas, other seasons primarily receive air masses from the northern region (Fig. 7). Aerosol chemical composition analysis in different clusters shows that summer air masses contain substantially higher proportions of SOA in PM_2.5 compared to other seasons (Fig. S4). This compositional difference leads to elevated NF_LH values in summer, resulting in significantly lower κ_mean values for accumulation-mode particles relative to winter and spring (Fig. 8).

Figure 8 The κ_mean (color block) of aerosols and the number fractions of different hygroscopic groups for all size particles (40–200 nm) in the influence of different air masses during different seasons. The color blocks of pie chart denote number fractions of different hygroscopic groups, i.e., orange block of pie chart for NF_MH, blue block of pie chart for NF_LH, and gray block of pie chart for NF_NH.

In contrast, autumn air masses exhibit higher POA content in PM_2.5 (Fig. S4), yielding relatively increased NF_NH and NF_LH compared to summer conditions, along with slightly lower κ_mean values for equivalent particle sizes (Fig. 8). While winter air masses demonstrate higher κ_MH (Fig. S5), spring air masses contain greater amounts of SNA in PM_2.5 (Fig. S4) and the elevated NF_MH ultimately produces slightly higher κ_mean values compared to that in winter (Fig. 8). Collectively, these observations reveal that similar particle hygroscopicity under winter–spring air masses, and comparable hygroscopic properties under summer–autumn air masses (Fig. 8).

Figure 9 The CV_κ-PDF for all size particles (40–200 nm) in the influence of different air masses during different seasons.

As illustrated in Fig. S4, the size of the pie charts represents the PM_2.5 mass concentration in different clusters. Notably, springtime PM_2.5 mass concentrations show significant variation among air mass categories, with C2 exhibiting relatively lower mass concentrations than C1 and C3. As indicated in Table S4, the occurrence frequencies of NPF events during spring for C1, C2, and C3 are 11 %, 42 %, and 25 %, respectively, suggesting that NPF events in Nanjing spring are more likely under the cleaner conditions of C2, characterized by lower PM_2.5 mass concentrations. This pattern may stem from reduced condensation sinks and higher abundances of gaseous precursors favorable for NPF in northerly air masses (Gysel et al., 2007; Li et al., 2023).

3.3.2 Influence of regional transport on hygroscopicity of particles of different sizes

Across all seasons except autumn, 200 nm particles exhibit relatively weak variation in κ_mean across clusters. This phenomenon is due to relatively small differences in both NF of different hygroscopicity groups and κ_MH across clusters within the same season (Figs. 8 and S5). In contrast, 200 nm particles during autumn display relatively large κ_mean variations across clusters, attributable to relatively larger differences in NF of different hygroscopicity groups despite comparable κ_MH values between air masses. These findings suggest that for 200 nm particles, NF of different hygroscopicity groups serves as the dominant factor governing κ_mean, a conclusion that aligns with the results presented in Sect. 3.1.2.

Integrative examination of Figs. 7, 8, and S4 further reveals that, within a given season, 200 nm particles originating from eastern/northern air masses consistently exhibit stronger hygroscopicity compared to those from western/southern sources. This enhanced hygroscopic behavior can be attributed to their higher SNA content in PM_2.5. This phenomenon also exists to some extent in accumulation-mode particles of other sizes, except during summer.

For 40 nm particles, κ_mean is predominantly influenced by NF_NH and NF_MH (Fig. 3), with NF_NH being relatively higher and NF_MH lower, particularly in autumn. Given that κ_NH remains relatively small and stable, seasonal variations in κ_mean are negligible despite relatively obvious NF_NH differences across air masses (Fig. 8). This implies that the hygroscopicity of 40 nm particles is less sensitive to air mass origins and may be primarily regulated by local sources. However, an exception occurs in spring, where air masses yield relatively higher κ_mean values for 40 nm particles, a consequence of their relatively elevated NF_MH (Fig. 8). However, for 80 nm particles, which are also in the Aitken mode but closer in size to the accumulation mode, the hygroscopicity differences influenced by air masses are similar to those of accumulation-mode particles and therefore are relatively significant across seasons (Fig. 8).

Figure 9 shows the CV_κ-PDF for size-resolved particles under varying air mass influences across seasons. Comparative analysis reveals that during non-summer seasons, 200 nm particles affected by long-range transported air masses (winter C3, spring C2, summer C2, autumn C3) demonstrate relatively higher CV_κ-PDF values relative to other air masses, suggesting relatively pronounced external mixing. Interestingly, summer exhibits a distinct pattern where C2-influenced 200 nm particles show even relatively lower CV_κ-PDF values compared to summer C1 and C3, indicating relatively advanced internal mixing. This seasonal anomaly may be attributed to enhanced photochemical aging during summer transport, driven by more intensive solar radiation. Consequently, while C2-associated 200 nm particles display marginally increased hygroscopicity during summer (Fig. 8c), other seasons show no statistically discernible hygroscopicity enhancement in long-range transported particle clusters.

All these results reveal relatively smaller intra-seasonal variations in air mass hygroscopicity compared to inter-seasonal differences. This pattern primarily stems from relatively pronounced chemical composition disparities between seasonal air masses. The study demonstrates distinct spatial controls on particle hygroscopicity: 40 nm particles are likely predominantly influenced by local sources, as evidenced by relatively minimal κ_mean variations despite relatively obvious NF fluctuations. In contrast, within a given season, 200 nm particles originating from eastern/northern air masses display relatively enhanced hygroscopicity relative to their western/southern counterparts, attributable to relatively elevated SNA concentrations. Notably, long-range air mass transport induces hygroscopicity slight enhancement exclusively during summer months for 200 nm particles, a phenomenon driven by accelerated aging processes and relatively advanced internal mixing. No comparable effects are detected in other seasons.

4 Summary and conclusions

This study examines size-resolved and seasonal variations in aerosol hygroscopicity and their influencing factors in northern Nanjing, based on field observations at NUIST in 2021 using H-TDMA measurements.

Aerosols in this suburban region exhibit relatively low hygroscopicity, which may be mainly due to relatively higher organic content (annual average MF_OA in PM_2.5: 42.92 %). The κ_mean increases with particle size across all seasons. Annual κ_mean values (± standard deviations) for 40, 80, 110, 150, and 200 nm particles are 0.12 ± 0.04, 0.16 ± 0.05, 0.18 ± 0.05, 0.20 ± 0.05, and 0.22 ± 0.06, respectively.

Winter and spring show relatively higher κ_mean as a result of the rise in MF_SNA under PM_2.5 pollution, which drives concurrent increases in both κ_MH and NF_MH. In contrast, summer and autumn exhibit relatively lower κ_mean. In winter, a high NO${}_{\mathrm{3}}^{-}$ content boosts κ_MH, especially in accumulation-mode particles. However, κ_mean is highest in spring due to a relatively increased SNA content, which drives the enhancement of NF_MH. Notably, 40 nm particles in spring show the highest κ_mean due to both increased NF_MH and decreased NF_NH, although the seasonal differences in κ_mean for 40 nm particles during non-spring periods are relatively small.

Across all seasons, κ_mean is mainly governed by NF_MH. For relatively smaller particles, κ_mean is also influenced by NF_NH, κ_MH, and κ_LH – particularly at 40 nm – while relatively larger particles are more affected by κ_MH. In contrast to winter and spring, κ_MH 's influence diminishes relatively in summer and autumn, while that of κ_LH correspondingly increases. Photochemical processes increase SO${}_{\mathrm{4}}^{\mathrm{2}-}$ and SOA during daytime, while NO${}_{\mathrm{3}}^{-}$ accumulates at night via aqueous-phase reactions (except winter). As a result, 200 nm particles show afternoon and evening κ_mean slight enhancement. Furthermore, traffic emissions cause κ_mean slightly dips across most sizes during morning and evening rush hours.

Spring records the highest frequency of new particle formation (NPF) events (21 %). During these events, newly formed particles are likely organic-rich, leading to relatively lower κ_mean for 40 nm particles originating from NPF in the morning. With the aging of particles, the κ_mean for 200 nm particles slightly enhanced in the late afternoon. Compared to winter, Aitken-mode particles on spring NPF days and spring non-NPF days show relatively higher κ_mean, while accumulation-mode particles are similar on spring non-NPF days but relatively lower than spring NPF days (except for 110 nm). Compared to spring non-NPF days, κ_mean on spring NPF days decreases (by 6.76 %) for 40 nm particles, and increases (by 5.07 %) for 200 nm particles.

Intra-seasonal variations in air mass hygroscopicity are relatively smaller than inter-seasonal differences, which may be attributed to the greater chemical contrast among seasonal air masses. Spatial controls differ by particle size: 40 nm particles may be primarily influenced by local sources, as indicated by relatively stable κ_mean despite NF variability. In contrast, within a given season, 200 nm particles transported from the east or north show relatively enhanced hygroscopicity due to higher SNA levels. Notably, long-range air mass transport enhances κ_mean for 200 nm particles only in summer, likely due to accelerated aging and more uniform internal mixing – an effect not observed in other seasons.

These findings provide valuable insights into the complex interactions between aerosol chemical composition, particle size, and seasonal-regional air mass transport in shaping aerosol hygroscopicity. Understanding these relationships is essential for improving the accuracy of regional climate models, particularly in estimating aerosol–cloud interactions and radiative forcing. Moreover, the study highlights the critical role of local emissions and secondary processes in influencing aerosol properties, offering a scientific basis for air quality management and pollution control strategies in rapidly urbanizing regions like Nanjing and the broader Yangtze River Delta.