Aerosol sampling in Xi'an took place during 1–30 April 2021 (spring), 12 July to 14 August 2021 (summer), 7 October to 7 November 2021 (fall) and 26 November to 31 December 2020 (winter). Xi'an has a population of ∼13 million and is located in the middle of the Guanzhong Plain, which is surrounded by the Qin Mountains and the Chinese Loess Plateau (Fig. S1 in the Supplement), favoring accumulation of air pollutants and formation of severe air pollution (Cao and Cui, 2021). In addition, Xi'an is adjacent to major deserts in China and is thus frequently affected by desert-dust aerosol.

Sampling in winter took place at an urban site (34.23∘ N, 108.89∘ E), which is close to a busy major road and is located in residential and commercial areas (Cao et al., 2012), and was carried out on a building roof (∼10 m from the ground) at the Institute of the Earth's Environment, Chinese Academy of Sciences. Sampling in the other three seasons took place at another urban site (34.37∘ N, 108.97∘ E), which is located in residential areas (Chen et al., 2021), and was carried out on a building roof (∼40 m from the ground) at Shaanxi University of Science and Technology. Meteorological parameters (wind speed and direction, temperature and relative humidity) and PM2.5 and PM10 mass concentrations were provided by nearby environmental monitoring stations.

Coarse (>1 µm) and fine (<1 µm) aerosol particles were collected onto Whatman 41 (W41) cellulose filters on a daily basis (from 08:00 to 07:30 the next day) using a two-stage aerosol sampler (TH-150C, Tianhong Co., Wuhan, China) with a flow rate of 100 L min-1. W41 filters, which were used for aerosol sampling, were acid-washed to reduce background levels. After aerosol collection, filters were sealed individually in clean plastic Petri dishes and then stored at -20 ∘C for further analysis. Our previous work (Zhang et al., 2022) described filter pretreatment, aerosol sampling and filter storage in detail. In the present work, 28, 32, 30 and 36 pairs of filter samples were collected in spring, summer, fall and winter, respectively.

In our work, mass concentrations of various species in air, including PM2.5 and PM10 concentrations, are reported under standard state conditions (at 0 ∘C and 1 atm) to remove the effects of variations in temperature and atmospheric pressure.

Sample analysis was detailed in our previous work (Zhang et al., 2022; Li et al., 2023), and as a result, here we only provide key information in brief. Every filter was equally halved. The first half-filter, which was used to determine total Fe, was digested in a Teflon jar using microwave digestion; after residual acids used in digestion were evaporated, the Teflon jar was cooled to room temperature and then filled with 20 mL HNO3 (1 %). The solution was filtered using a polytetrafluoroethylene (PTFE) membrane syringe filter (pore size: 0.22 µm) and then analyzed using inductively coupled plasma mass spectrometry (iCAP Q, Thermo Fisher Scientific, USA). In total, 14 elements were determined, including Fe, Al and Pb, and the recovery rates were found to be 90 %–110 % for Fe using certificated reference materials (GBW07454 and GSB07-3272-2015).

The other half-filter, which was used to determine dissolved Fe and soluble ions, was immersed in 20 mL ultrapure water for 2 h, during which an orbital shaker (300 r min-1) was used to stir the aqueous mixture. After that, the aqueous mixture was filtered using a PTFE membrane syringe filter (pore size: 0.22 µm) and then divided further into two parts. The first solution (∼10 mL) was analyzed by ion chromatography to measure soluble anions and cations; the second solution (10 mL) was acidified to contain 1 % HNO3 (using 147 µL 67 % HNO3) and subsequently analyzed using inductively coupled plasma mass spectrometry.

The ISORROPIA-II model (Fountoukis and Nenes, 2007) was used in the “metastable + forward” mode to calculate aerosol pH for coarse and fine particles, and input data included concentrations of soluble anions (SO42-, NO3- and Cl-) and cations (NH4+, Na+, K+, Ca2+ and Mg2+), temperature and relative humidity (RH). The effects of NH3(g) and HNO3(g) were not taken into account as their concentrations were not measured; this may cause some biases (likely underestimation) in calculated aerosol pH (Guo et al., 2015; Hennigan et al., 2015; Pye et al., 2020), but the overall trend of aerosol pH would not be significantly affected. The reverse mode was not used in our work, as results calculated using the reverse mode are very sensitive to uncertainties in concentrations of common aerosol ions (Hennigan et al., 2015). Coarse particles are generally expected to be less acidic than fine particles, and it is not clear yet why similar and even lower aerosol pH was observed for coarse particles (when compared to fine particles) in spring and fall at Xi'an (Fig. S14). This may be caused by biases in aerosol pH calculation, and lower aerosol pH values were also reported in previous work for coarser particles carried out in northern Colorado, United States (Young et al., 2013). Concurrent measurements of gaseous NH3, HNO3 and HCl would help reduce uncertainties in calculated aerosol pH and have been implemented in our following studies since July 2022.

The HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory)-4 model (Draxier and Hess, 1998) was employed to calculate 48 h air mass back trajectories using meteorological data (horizontal resolution: 1∘×1∘; time resolution: 3 h) from the Global Data Assimilation System provided by the National Centers for Environmental Prediction. Back trajectories were determined with an arrival height of 100 m a.g.l. and arrival times of 08:00, 14:00, 20:00 and 02:00 on the next day (Wang et al., 2020), and every day four back trajectories were obtained. In total, 120, 136, 128 and 144 back trajectories were obtained in our work for spring, summer, fall and winter, and all the back trajectories were clustered using the cluster analysis method described elsewhere (Baker, 2010).

The climate in Xi'an (and the Guangzhou Plain in general) is mainly regulated by the East Asian monsoon. During our campaign, the prevailing wind directions were west and southwest in spring, northeast in summer, southwest and northeast in fall, and west in winter (Fig. S2); furthermore, average wind speeds were >2 m s-1 in summer and fall and <2 m s-1 in spring and winter. Median temperatures were 13.6, 27.0, 12.7 and 1.3 ∘C in spring, summer, fall and winter, and median RH was found to be 85 %, 71 %, 83 % and 77 % (Table S1 in the Supplement). Precipitation mainly took place in summer during our campaign in 2021, similarly to previous years (Cao and Cui, 2021).

Table S2 shows PM2.5 and PM10 concentrations at Xi'an in the four seasons. PM10 concentrations were in the ranges of 15–243, 24–76, 22–151 and 41–212 µg m-3 in spring, summer, fall and winter, and the average values were 93 ± 61, 51 ± 16, 70 ± 35 and 107 ± 39 µg m-3, suggesting the highest levels in spring and winter and the lowest levels in summer. PM2.5 mass concentrations were in the ranges of 11–62, 11–48, 13–97 and 13–156 µg m-3, and the average values were 35 ± 14, 23 ± 8, 40 ± 24 and 80 ± 32 µg m-3, suggesting the highest concentrations in winter and the lowest levels in summer.

PM2.5 and PM10 concentrations were high in winter due to accumulation of anthropogenic pollution, and the median PM2.5 / PM10 ratio was 0.76. PM10 concentrations in spring were significantly increased due to the impacts of desert-dust aerosol, and the median PM2.5 / PM10 ratio was only 0.44. Two major dust events occurred during our spring campaign (12–17 and 27–30 April). During the two dust events the average PM10 and PM2.5 mass concentrations were 151 ± 57 and 42 ± 12 µg m-3, and PM2.5 / PM10 ratios became even lower (0.20–0.31). Furthermore, the median PM2.5 / PM10 ratios were 0.44 in summer and 0.62 in fall.

Figure 1a shows the seasonal variation of total aerosol Fe in coarse and fine particles measured by our work at Xi'an. Total aerosol Fe concentrations were in the ranges of 270–3095, 191–1992, 395–3492 and 269–3924 ng m-3 for coarse particles in spring, summer, fall and winter, and the average values were 1504 ± 800, 950 ± 524, 1638 ± 830 and 1831 ± 866 ng m-3 (Table A1 in the Appendix); total aerosol Fe concentrations were in the ranges of 206–12 144, 164–1591, 196–2631 and 257–4268 ng m-3 for fine particles in spring, summer, fall and winter, and the average values were 3717 ± 3387, 721 ± 366, 958 ± 516 and 2058 ± 1037 ng m-3. Average total Fe concentrations were measured to be 798 ± 466 and 801 ± 534 ng m-3 for coarse and fine particles in winter (November–December 2019) at Qingdao (Zhang et al., 2022), only 44 % and 38 % of the average values (1831 ± 866 and 2058 ± 1037 ng m-3) found in winter (November–December 2020) at Xi'an by the present work, mainly because during wintertime aerosol mass concentrations were much higher at Xi'an than Qingdao.

The average contributions of coarse particles to total Fe in TSPs (total suspended particles) were 29 %, 57 %, 63 % and 47 % in spring, summer, fall and winter, being lowest in spring, when the influence of desert-dust aerosol was largest. Statistical analysis (paired t test) suggested that, compared to fine particles, total Fe in coarse particles was significantly lower in spring (p<0.01, α=0.05) but significantly higher in summer and fall (p<0.01, α=0.05); in addition, there was no significant difference between coarse and fine particles in winter (p=0.13, α=0.05). The average contributions of coarse particles to total Al in TSPs were 26 %, 50 %, 54 % and 40 % in spring, summer, fall and winter, also being lowest in spring. Furthermore, during dust periods (12–17 and 27–30 April), the average contribution of coarse particles was found to be 24 % for total Fe and 23 % for total Al in TSPs, even slightly smaller than the average values in spring. This is probably because, for desert-dust aerosol, Fe and Al were enriched in fine particles, while the major component in coarse particles was quartz (Journet et al., 2014); nevertheless, further measurements are needed to better understand the size distribution of trace metals in desert-dust aerosol.

Compared to other seasons, total Fe concentrations in fine and coarse particles were lowest in summer (Fig. 1a), as aerosol mass concentrations were also lowest in summer (Sect. 3.1). Similarly, previous measurements on Huaniao Island in the East China Sea (Yang et al., 2020) and over the tropical and subtropical North Atlantic (Chen and Siefert, 2004) also found the lowest total aerosol Fe levels in summer. For the other three seasons (spring, fall and winter), the total Fe in coarse particles was rather similar, while the total Fe in fine particles was highest in spring and lowest in fall. Overall, compared to summer and fall, the total aerosol Fe was higher in spring and winter, when higher aerosol mass concentrations were also observed (Fig. S3).

Total Fe was very well correlated with total Al (0.87<r<0.96, p<0.01) for both coarse (Fig. S4) and fine particles (Fig. 2) in all four seasons, suggesting desert dust always as the dominant source of total aerosol Fe at Xi'an, regardless of particle size range and seasons. The median Fe/Al values, the mass ratios of total Fe to total Al, were 0.975, 0.926, 1.269 and 0.940 in spring, summer, fall and winter for coarse particles and 0.735, 0.796, 0.870 and 0.744 for fine particles (Fig. S5). Fe/Al was found to be 0.911 and 0.741 for PM10 and PM2.5 generated using surface soil samples collected over several major deserts in China (Zhang et al., 2014). We found that Fe/Al measured for coarse and fine particles at Xi'an was similar to that reported for desert dust (Zhang et al., 2014); coarse particles in fall might be one exception (Fig. S5), showing slightly higher Fe/Al (median: 1.269) than desert dust.

Figure 1b shows the seasonal variation of dissolved aerosol Fe in coarse and fine particles at Xi'an. Dissolved aerosol Fe concentrations were in the ranges of 1.0–17.5, 0.9–22.0, 0.7–42.2 and 0.9–33.8 ng m-3 for coarse particles in spring, summer, fall and winter, and the average values were 5.9 ± 4.5, 5.6 ± 4.0, 13.5 ± 12.2 and 14.5 ± 8.3 ng m-3; dissolved aerosol Fe concentrations were in the ranges of 1.1–21.4, 2.6–22.9, 2.1–33.7 and 5.0–89.5 ng m-3 for fine particles in spring, summer, fall and winter, and the average values were 10.0 ± 5.5, 9.7 ± 5.6, 16.5 ± 10.1 and 22.7 ± 16.8 ng m-3. Average dissolved Fe concentrations were measured as 7.7 ± 14.5 and 7.3 ± 7.6 ng m-3 for coarse and fine particles in winter at Qingdao (Zhang et al., 2022), only 53 % and 32 % of the average values (14.5 ± 8.3 and 22.7 ± 16.8 ng m-3) found in winter at Xi'an by the present work, and one major reason is that total Fe concentrations were significantly higher at Xi'an than Qingdao.

The average contributions of coarse particles to dissolved Fe in TSP were 37 %, 36 %, 45 % and 39 % in spring, summer, fall and winter. Compared to fine particles, dissolved Fe was significantly lower in coarse particles for all four seasons (paired t test, p<0.01, α=0.05) at Xi'an, although the total Fe in coarse particles was higher than or similar to fine particles (except spring, as discussed in Sect. 3.2). This indicated that aerosol Fe solubility was lower in coarse particles than fine particles, as further discussed in Sect. 4. Similar to our work, Sakata et al. (2022) found that, over the Pacific, dissolved aerosol Fe concentrations in fine particles (<1.3 µm) were significantly higher than coarse particles (>1.3 µm).

Compared to spring and summer, dissolved Fe concentrations were higher in fall and winter for coarse particles (Fig. 1b); for fine particles, dissolved Fe concentrations were highest in winter, followed by fall, and lowest in spring and summer. Dissolved Fe concentrations were low in summer, as total Fe concentrations were also low (Fig. 1a). Total Fe concentrations were high in spring (Fig. 1a), but dissolved Fe concentrations were low; this is because, compared to other seasons, spring was most severely affected by desert dust with low Fe solubility. Our previous study (Zhang et al., 2022) investigated aerosol Fe at Qingdao in winter and found that, compared to clean days, dissolved Fe concentrations did not change significantly during dust days, although total Fe concentrations were remarkably increased. Therefore, our previous (Zhang et al., 2022) and current studies imply that the occurrence of desert-dust aerosol may not necessarily lead to an increase in dissolved Fe concentrations in the air.

As shown in Fig. S6, overall the correlation between dissolved Fe and total Al was quite weak at Xi'an, indicating that desert dust may not contribute dominantly to dissolved aerosol Fe, although it was always the major source of total aerosol Fe (Sect. 3.2). We also examined correlations between dissolved Fe and several other species (Table S3). Except for summer, dissolved Fe was well correlated with secondary inorganic species (sulfate, nitrate and ammonium) for coarse and fine particles, suggesting secondary formation (i.e., conversion of insoluble Fe to dissolved Fe via chemical processing) as an important source of dissolved Fe. In addition, Fig. 3 shows that dissolved Fe was well correlated with K+ (a tracer for biomass burning) in coarse and fine particles in all four seasons (0.67<r<0.96, p<0.001), and this may indicate biomass burning also as an important source of dissolved aerosol Fe.

Furthermore, good correlations with dissolved Fe were found in coarse particles for Pb, Zn and As in three seasons (spring, fall and winter) and in fine particles for Pb (spring, fall and winter), As (spring and fall) and Zn (fall). Aerosol Pb and Zn are mainly emitted by vehicles and the iron–steel industry (Chow et al., 2004; Cao and Cui, 2021), and the major sources of aerosol As include coal combustion and metal smelting (Tian et al., 2010). As a result, vehicle emission, coal combustion, the iron–steel industry and metal smelting also contributed to dissolved aerosol Fe at Xi'an.

Back trajectories obtained for our campaign were clustered, and we further examined the dependence of Fe solubility on air mass cluster types in different seasons. In spring (Fig. 6a), air mass cluster C1 originated locally, and C2 originated from the North China Plain with severe air pollution, while C3 and C4 represented air mass arriving from desert regions in the north and northwest after long-distance transport (compared to C1 and C2); as shown in Fig. 6b, Fe solubility in coarse and fine particles was significantly higher for C1 and C2 when compared to C3 and C4.

In fall (Fig. 6e), air mass cluster C1 originated locally, C2 was transported from desert regions in the north and northwest, and Fe solubility was much higher for C1 than C2 (Fig. 6f). In winter (Fig. 6g), air mass cluster C1 originated locally, while C2, C3 and C4 originated from desert regions in the north and northwest, and the transport distance increased from C2 to C4; as shown in Fig. 6h, Fe solubility followed the order C1 > C2 > C3 > C4, decreasing with an increase in transport distance. In contrast to the other three seasons, no obvious dependence of Fe solubility on air mass clusters was found in summer (Fig. 6d).

To summarize, our work found that, in spring, fall and winter, Fe solubility at Xi'an was significantly higher when air masses originated from local and nearby regions when compared to those arriving from desert regions after long-distance transport. The reason is that the contribution of anthropogenic emissions to aerosol Fe was elevated for air masses originating from local and nearby sources (when compared to air masses originating from desert regions), and anthropogenic aerosol Fe had higher solubility than desert dust (Schroth et al., 2009; Fu et al., 2012; Oakes et al., 2012b). Similar to our work, over the Sargasso Sea aerosol Fe solubility was much lower in Saharan air masses than North American air masses (Sedwick et al., 2007).

Laboratory studies (Shi et al., 2011a; Chen and Grassian, 2013; Wang et al., 2018) suggested that chemical processing by acids, such as H2SO4 and HNO3, could dissolve insoluble Fe and thus enhance aerosol Fe solubility. Some field studies found that aerosol Fe solubility was positively correlated with sulfate and/or nitrate (Shi et al., 2020; Zhu et al., 2020; Liu et al., 2021; Zhang et al., 2022; Yang et al., 2023), indicating enhancement of Fe solubility by atmospheric acid processing.

Figure 7 plots Fe solubility at Xi'an versus (2 × [sulfate] + [nitrate]) / [Fe] (in nanomole per nanomole, referred to as relative abundance of aerosol acidic species), the molar ratio of two major acidic species to total Fe in aerosol particles. For coarse particles, aerosol Fe solubility was well correlated with relative abundance of aerosol acidic species in all four seasons (0.77<r<0.91, p<0.01). For fine particles, good correlation was found in spring (r=0.84, p<0.01), moderate correlation was found in fall and winter (0.42<r<0.53, p<0.01), and no significant correlation was found in summer (r=0.35, p>0.05). In addition, as shown in Figs. S10–S11, correlations of Fe solubility with [nitrate] / [Fe] were better than (or very similar to) those with [sulfate] / [Fe] for coarse particles in the four seasons, whereas no obvious trend was observed for fine particles.

Overall, correlations between Fe solubility and relative abundance of aerosol acidic species were always better for coarse particles than fine particles (Fig. 7), indicating that acid processing may be more important in Fe solubility enhancement for coarse particles when compared to fine particles. A previous study (Zhang et al., 2022) also found that such correlation was better in coarse particles than fine particles in winter at Qingdao, a coastal city in northern China. Nevertheless, as discussed in Sect. 4.1, Fe solubility was higher in fine particles than coarse particles. This may imply that primary emission of non-desert-dust Fe (anthropogenic Fe) with higher solubility (Schroth et al., 2009; Oakes et al., 2012b) was more important for Fe solubility enhancement in fine particles than coarse particles.

It was suggested by laboratory studies (Chen and Grassian, 2013; Paris and Desboeufs, 2013; Wang et al., 2017) that atmospheric organic ligands, such as oxalate, could increase aerosol Fe solubility via ligand-promoted dissolution. As shown in Fig. S12, our present work found good correlation between Fe solubility with [oxalate] / [Fe] (in nanomole per nanomole) for coarse particles (0.70<r<0.88, p<0.01) and moderate correlation for fine particles (0.40<r<0.67, p<0.01) at Xi'an. Positive correlation between Fe solubility and oxalate was also observed previously in Atlanta, USA (Yang and Weber, 2022), Toronto, Canada (Tao et al., 2022), and Qingdao, China (Zhang et al., 2022).

We note that good correlation between Fe solubility and aerosol oxalate does not necessarily mean Fe solubility enhancement by Fe–oxalate complexation. For example, it was suggested that Fe could promote the formation of oxalate in aerosol particles (Tao and Murphy, 2019; Zhang et al., 2019), and thus good correlation between Fe solubility and oxalate could also imply enhanced formation of oxalate by dissolved Fe. In addition, similar to sulfate and nitrate, the major source of oxalate in the troposphere was secondary formation (Myriokefalitakis et al., 2011; Kawamura and Bikkina, 2016), and in this aspect good correlation between Fe solubility and relative abundance of oxalate could also indicate the importance of secondary formation of dissolved aerosol Fe (i.e., dissolution of insoluble Fe to dissolved Fe via aging processes).