Wuhan (30.5° N, 114.4° E) is a major industrial city in central China with a population of over 13 million, producing abundant urban anthropogenic emissions from sources such as vehicle exhaust and industrial fossil fuel combustion (Zhang et al., 2015a). The city lies in the subtropical monsoon region and experiences four distinctive seasons. In winter, the northeast monsoon brings cold, relatively dry weather (Wu and Wang, 2002). Frequent temperature inversions in the lower troposphere suppress vertical convection and often lead to severe air pollution (Zhang et al., 2021). In summer, the southeast and southwest monsoons cause high temperatures and heavy rainfall, creating favorable conditions for pollution dispersion (Ding et al., 2015). Spring and autumn are the transition phases between these two regimes. Additionally, regional air mass transport also plays a significant role in aerosol loading over Wuhan. In spring and winter, mineral dust from deserts in north and northwest China as well as in Mongolia is frequently long-range transported to Wuhan (He and Yi, 2015; Jing et al., 2024). In summer and autumn, agricultural biomass burning in neighboring provinces sometimes contributes to the poor air quality and severe haze events over Wuhan (Zhang et al., 2014; Jing et al., 2025).

Among the major aerosol types in Wuhan, mineral dust is generally considered hydrophobic. In contrast, all the non-dust components are classified as anthropogenic aerosols, which exhibit varying degrees of hygroscopicity due to the inclusion of water-soluble inorganic ions and organic matter. Our lidar site is located in central Wuhan, an area surrounded by an extensive water network (the Yangtze River and numerous urban lakes; Fig. 1), which creates a humid atmospheric environment that facilitates aerosol water uptake. Therefore, particle hygroscopic growth may contribute to the lidar-derived particle backscatter coefficients. In this study, all lidar-derived aerosol optical properties discussed, including the particle backscatter coefficient, extinction coefficient, and AOD, are attributed exclusively to anthropogenic aerosols.

Height-resolved aerosol optical properties over Wuhan have been observed using a 532 nm ground-based polarization lidar since October 2010 (Yin et al., 2021; He et al., 2024; Jing et al., 2024, 2025). Detailed specifications of the lidar system were provided in Kong and Yi (2015). In 2017, a transparent waterproof window was installed on top of the lidar container, enabling uninterrupted lidar operations regardless of rainy or snowy weather conditions (Yi et al., 2021). Raw data are stored with a time resolution of 1 min and a vertical resolution of 30 m. The lowermost height with complete field-of-view (FOV) observation is 0.3 km. Specifications of the polarization lidar system are listed in Table 1.

The volume depolarization ratio δv (VDR) is defined as the ratio of perpendicular- to parallel-oriented signals, multiplied by the gain ratio between the two polarized channels, and is used to derive the particle depolarization ratio δp (PDR) (Freudenthaler et al., 2009). The particle backscatter coefficient βp and particle extinction coefficient αp are retrieved using the Fernald method (Fernald, 1984), assuming a fixed lidar ratio of 50 sr (Wang et al., 2016). In addition, the non-dust particle backscatter coefficient βnd is calculated using the polarization-lidar photometer networking (POLIPHON) method as follows (Tesche et al., 2009; Mamouri and Ansmann, 2014): 1βndz=βpz-βpzδpz-δnd1+δdδd-δnd1+δpz where z represents the altitude; δd=0.31 and δnd=0.05 are the particle depolarization ratios for dust and non-dust, respectively. For each profile, we set βndz=βpz if δpz<δnd and βndz=0 if δpz>δd (Mamouri and Ansmann, 2014). As noted in the previous section, the non-dust component primarily corresponds to anthropogenic aerosols over Wuhan; thus, βnd and αnd represent the optical properties of anthropogenic aerosols. Table 2 lists the uncertainties in the lidar-derived aerosol optical property parameters.

Cloud-free profiles with signal accumulation times of 30–80 min are obtained using a cloud screening algorithm (Yin et al., 2021). For each cloud-free profile, the particle backscatter coefficient (i.e., total, dust, and non-dust components), and both volume and particle depolarization ratios are retrieved. We utilize the same methodology as Yin et al. (2021) and extend the aerosol profiles to September 2024 (Jing et al., 2025). A total of 24 910 cloud-free profiles are identified from 2139 observational days between October 2010 and September 2024. To avoid potential contamination from severe haze or fog, cloud-free profiles containing one or more vertical bins with extinction coefficients exceeding 1.5 km⁻¹ are excluded. In total, 676 (2.7 % of all) cloud-free profiles are removed from the analysis. This mature dataset is used to further investigate hygroscopic growth.

The Hybrid Single Particle Lagrangian Integrated Trajectory (HYSPLIT), developed by the National Oceanic and Atmospheric Administration Air Resources Laboratory (NOAA ARL), was used to simulate both forward and backward air mass trajectories (HYSPLIT, 2025). These simulations are driven by meteorological field data from the GDAS archive (Kanamitsu, 1989) and require initialization parameters such as start time, altitude, and geographical location (Draxler and Rolph, 2003; Stein et al., 2015). In this study, three backward trajectories arriving at Wuhan at different altitudes were simulated to trace the potential origins of aerosols at those altitudes.

Two radiosonde launches were conducted daily at 08:00 local time (LT) and 20:00 LT, at 30.6° N, 114.1° E, approximately 24 km away from our lidar site. The sondes measured vertical profiles of temperature, pressure, relative humidity (RH), water vapor mixing ratio, and wind speed/direction from the surface to up to ∼ 30 km altitude. The measurement error for temperature is less than 1 °C, and the uncertainty in RH is below 5 % when the temperature exceeds -10 °C (Nash et al., 2011). The potential temperature θ is defined as (Bolton, 1980): 2θ=T⋅P0P0.286 where T is the temperature (K), P is the atmospheric pressure (hPa), and P0 is the reference pressure of 1000 hPa. In this study, radiosonde data were interpolated to match the corresponding altitude bins of the lidar profiles using a cubic spline interpolation method. This interpolation ensured consistent alignment between radiosonde and lidar measurements, facilitating our analysis of aerosol hygroscopic growth.

The European Center for Medium-Range Weather Forecasts (ECMWF) reanalysis version 5 (ERA5) (Copernicus Climate Change Service, 2025) provides global atmospheric reanalysis data from January 1940 onward. ERA5 combines model outputs with worldwide observations into a globally consistent and physically constrained dataset (Hersbach et al., 2020). It offers hourly estimates of atmospheric, land, and oceanic climate variables. The boundary layer height (BLH) is the depth of the boundary layer (BL) directly affected by dynamic, thermal, and other surface interactions (Peng et al., 2023). Above the BL is the free troposphere (FT), where aerosols primarily originate from non-local sources (Bourgeois et al., 2018). In this study, ERA5 hourly BLH data (Hersbach et al., 2023) for Wuhan were used to distinguish respective hygroscopic growth cases occurring in the BL and FT. Considering the presence of an aerosol residual layer, the diurnal maximum BLH from ERA5 was adopted as the boundary between the BL and FT.

Figure 5 presents the probability density distribution of the particle backscatter coefficient enhancement factor fref(RH) for 192 identified cases from lidar observations during 2010–2024. The average R2 value of 0.91 indicates a good fit with the Hänel parameterization for most cases. The average γ value was 0.62 ± 0.24, represented by the blue curve and shaded area, corresponding to fref-Hänel(85%) of 2.36, with a range of 1.69–3.29 when including the standard deviation. The γ value in the BL (0.63 ± 0.25) was comparable to that in the FT (0.60 ± 0.24). The mean RH in the lower and middle troposphere was 71.2 ± 12.5 % (68.4 ± 12.0 % in the BL and 75.3 ± 12.1 % in the FT), corresponding to the highest probability density area (in yellow). This suggests that most hygroscopic growth of urban anthropogenic aerosols in Wuhan occurred under high RH conditions of around 60 %–80 %. The particle backscatter coefficient enhancement factor fref-Hänel(85%) exhibited a broad distribution from 1.32 to 4.53, corresponding to γ values of 0.20–1.09, likely reflecting the diverse hygroscopicity properties of urban particles in Wuhan. Rapid urbanization over the past decades has exposed the city to numerous pollutants, including various water-soluble inorganic ions, elemental carbon, and organic matter, etc. (Zhang et al., 2015a). Both chemical composition and particle size significantly influence aerosol hygroscopicity characteristics (Zieger et al., 2010, 2013), resulting in a variety of hygroscopic aerosol types.

In the previous section, dust and anthropogenic aerosols are assumed to be externally mixed. However, the internally mixing conditions are unavoidable for East Asian dust events (Xu et al., 2020), which may lead to misclassification of “coated dust” as “anthropogenic” by the POLIPHON method. To assess the potential influence of internally mixed dust, a sensitivity analysis was conducted. Given that the δp during dust events over Wuhan varies between 0.1 and 0.3 (Jing et al., 2024), the threshold value for pure dust δd (in Eq. 1) was reduced to lower values of 0.10–0.25, thereby making the extraction of “anthropogenic aerosols” more conservative. As a result, the γ increases only slightly by 0.02 (0.62 to 0.64). Moreover, approximately 54.7 % cases in this study show no dust interference, and for the remaining cases the dust optical depth (DOD) does not exceed 0.05. It can be concluded that the error introduced by potential misclassification is limited.

The annual mean γ from 2010–2024 is presented in Fig. 5d, with the number of identified cases for each year indicated at the top of each bar. The annual mean γ generally ranged from 0.5 to 0.7. Notably, the annual mean γ sharply increased from 0.49 in 2015 to 0.63 in 2017, and stabilized high between 0.6 and 0.7 after 2018. The evolution of the annual mean NO₂-to-SO₂ concentration ratio in Wuhan from 2014 to 2024 is also presented in figure 5d (red dashed broken line), which closely follows the trend of γ. The NO₂-to-SO₂ concentration ratio increased sharply from 1.8 in 2014 to 5.3 in 2017, and varied between 4 and 6 during 2018–2024, suggesting that the disparity in emission control measures for these two gaseous precursors, i.e., favoring more particulate nitrate formation, likely contributes to the increase in γ after 2017. Chen et al. (2019) reported that a higher nitrate fraction in an aerosol mixture enhances aerosol hygroscopicity under the same RH conditions. In addition, our previous study showed that the anthropogenic AOD at 532 nm over Wuhan declined during 2010–2017 due to strict emissions control policies, but this downward trend ceased from 2018 onwards (Jing et al., 2025). The relatively higher γ values post-2017 result in a larger backscatter coefficient enhancement factor, indicating that AOD values after 2017 contain more contributions from hygroscopic growth effect, which partly offsets the effectiveness of emission control policies.

Figure 6 presents the seasonal variation of the particle backscatter coefficient enhancement factors, derived from the 192 identified cases. Seasons are defined as spring (March–April–May), summer (June–July–August), autumn (September–October–November), and winter (December–January–February). Most cases occurred in summer (80) and autumn (81), as dust is commonly present in spring and winter (Jing et al., 2024) and was therefore excluded from the analysis. There is no significant seasonal difference in γ, with a maximum of 0.64 in autumn and a minimum of 0.56 in winter, corresponding to fref-Hänel(85%) values of 2.43 and 2.17, respectively, i.e., a difference of approximately 11 %. Urban pollutants emitted from human activities do not vary substantially with seasons, except for nitrate, which shows the highest fraction in PM_2.5 during winter (Zhang et al., 2015a).

To explain why the fraction of nitrate in Wuhan is the highest in winter, while γ is the lowest (0.56), the mean layer heights (MLH) for cases in the four seasons are presented in Fig. 6. The MLH in winter is 2.5 km, higher than the 1.4–1.5 km observed in other seasons, likely due to frequent dust intrusions below 1.5 km in winter (Jing et al., 2024). However, anthropogenic aerosols in winter are usually concentrated below 1.5 km (Jing et al., 2025). Therefore, the winter analyzed here represents hygroscopic growth effect at higher altitudes in the free troposphere, where the atmosphere is relatively clean, rather than surface-level pollution. Chen et al. (2019) reported an extremely low γ of 0.1 under clean conditions, revealing that the hygroscopicity of clean air is lower than that of polluted air. Similarly, Sicard et al. (2022) found only slight seasonal variation in γ values in Barcelona, with a maximum of 0.58 in summer and a minimum of 0.53 in autumn. They interpreted this season-independent γ as recirculation layers of pollutants above the BL, caused by strong insolation, weak synoptic forcing, sea breezes, and mountain-induced winds (Pérez et al., 2004).

Table 3 summarizes hygroscopic growth parameters from this study in Wuhan and from the existing literature measured elsewhere. A variety of aerosol mixtures was examined, with γ values varying from 0.24 for mixtures dominated by hydrophobic particles, such as dust, to 1.1 for highly hygroscopic aerosols (e.g., marine aerosols and inorganic salts). To analyze the causes of the lidar-observed aerosol hygroscopic growth, many studies indirectly inferred the aerosol composition through backward trajectories and optical properties (e.g., Ångström exponent, complex refractive index, depolarization ratio) (Veselovskii et al., 2009; Granados-Muñoz et al., 2015; Sicard et al., 2022; Haarig et al., 2025). Miri et al. (2024) introduced fluorescence capacity, which was not affected by water vapor, to distinguish aerosol components (with biological aerosols exhibiting higher fluorescence, and pure dust or urban aerosols demonstrating lower fluorescence). Ground-based aerosol chemical speciation monitor (ACSM) was also used to explain the hygroscopic growth behavior of aerosols through ground-level chemical composition analysis (Lv et al., 2017; Bedoya-Velásquez et al., 2018; Chen et al., 2019; Wu et al., 2020). Pérez-Ramirez et al. (2021) provided the first airborne in situ measurements for chemical composition determination, confirming that sulfates and water-soluble organic carbon are the main contributors to the aerosol hygroscopic growth observed by lidar, with γ values of 0.38–0.39. In addition, Laly et al. (2025) combined lidar-based hygroscopic growth estimation with chemical species data from the Copernicus atmospheric monitoring service (CAMS), showing that γ was significantly higher in regional pollution cases affected by sea salt (0.87 and 1.52) compared with those without sea salt influence (0.30–0.75).

The γ value from this study (0.62 ± 0.24) is slightly higher than that reported by Sicard et al. (2022) (0.55 ± 0.23), who conducted a statistical analysis of hygroscopic growth parameters for local/regional pollutants and sea salt aerosols without dust interference in Barcelona, Spain. Several factors may explain this difference. First, although sea salt was not considered in the present study, we speculate that the diverse inorganic salts emitted from extensive human activities in Wuhan probably contributed significantly to the observed strong hygroscopicity. Liu et al. (2014) found that the fractions of ammonium, nitrates, and sulfate are strongly correlated with aerosol hygroscopicity. Similarly, He et al. (2016) identified a region of high hygroscopic growth in Eastern China, corresponding to large-scale industrial districts with substantial emissions of inorganic salts, such as sulfates and nitrates. Wu et al. (2020) also measured a γ value of 1.14 for fine-mode inorganic salts in urban pollution over Beijing. Second, both Sicard et al. (2022) and this study are based on lidar observations within two hours before and after the radiosonde launches (00:00 and 12:00 UTC). During this period, Barcelona corresponds to noon or midnight (UTC+1 in winter and UTC+2 during daylight saving time). In contrast, Wuhan (UTC+8) experienced the morning and evening rush hours, which contribute substantially to traffic emitted NO₂. Zhang and Cao (2015b) found two NO₂ emission peaks in Chinese megacities (Beijing, Shanghai, and Guangzhou, which have traffic patterns similar to Wuhan) between 07:00–10:00 and 19:00–22:00 local time. As discussed previously, higher NO₂ emissions favor the formation of hygroscopic nitrate particles, contributing to the larger γ values observed in Wuhan.

In Sect. 2, the backscatter coefficient βp is retrieved by the Fernald method with a fixed lidar ratio (LR) of 50 sr (Fernald, 1984). However, the hygroscopic growth process can lead to an increase in LR under high RH conditions in previous studies: Veselovskii et al. (2025) found that during the hygroscopic growth, the extinction increases more rapidly than the backscatter; Haarig et al. (2025) estimated the LR enhancement factor of 1.43 when RH =90%. While the polarization lidar cannot measure the LR for specific cases like the Raman lidar. Zhao et al. (2017) have estimated a relational expression between LR and RH through 532 nm micro-pulsed lidar and Mie model: 7LR=LRdry×0.92+2.5×10-2RH-40-1.3×10-3RH-402+2.2×10-5RH-403 where LRdry represents the LR under dry conditions. The fixed LR of 50 sr represents an average value derived from combined lidar and sun photometer measurements in the ambient troposphere (Takamura et al., 1994). Based on radiosonde data, the average RH in the lower troposphere over Wuhan is approximately 40 %–70 % (Guo et al., 2023). Accordingly, LRdry=47 sr in Eq. (7) was set, such that an LR of 50 sr corresponds to RH values of approximately 50 %–55 %. It should be mentioned that our analysis focuses on the variation of LR within the identified particle-hygroscopic-growth layers. Taking the case from 19 July 2019, presented in Sect. 3 as an example, Fig. 7 illustrates the influence of variable LR on hygroscopic parameter γ. The LR increases with rising RH, and the derived backscatter coefficient (in red curve) shows a slight deviation from the original profile (in black curve). The derived γ increases by 2.1 % from 0.48 to 0.49. Furthermore, Table 4 summarizes 10 cases covering RH ranges of 40 %–100 %. The variable LR generally causes an increase in γ of <10%, with the largest increase up to 12.5 % under high RH conditions. The uncertainty introduced by assuming a fixed LR becomes more pronounced under higher RH conditions.