The WLG (36°17^′ N, 100°54^′ E; 3816 m a.s.l.) is situated at the summit of the Waliguan Mountain on the northeastern TP and is part of the Global Atmosphere Watch (GAW) program operated by the World Meteorological Organization (WMO) (Fig. 1). The Waliguan Mountain, with a relative elevation difference of about 600 m above the surrounding terrain (Fig. 1b), is an ideal location for studying the background characteristics of the atmospheric environment over inner Asia. The WLG is about 90 km west of Xining, the capital of Qinghai Province with an elevation of ∼ 2300 m a.s.l. The regional climate is characterized by distinct seasonal variations, strongly influenced by the Asian summer monsoon and East Asian winter monsoon. Spring and fall serve as transitional seasons between these two climatic systems.

Fine particulate matter (PM_2.5) filter samples were collected on 47 mm diameter quartz fiber filters (PALL Life Sciences, USA) using a low flow aerosol sampler (Wuhan Tianhong Instrument Co. LTD, TH-16E) operating at a flow rate of 16.7 L min⁻¹. Before sampling, the filters were baked in a muffle furnace at 550 °C for 4 h to remove residual organic material. After sampling, each filter was stored in a clean filter holder wrapped in aluminum foil and stored at -18 °C. A total of 48 filter samples and three field blanks were collected from 14 June 2019 to 6 May 2020. Each sample was collected for 48 h with a sampling frequency of once every seven days. Field blanks were obtained by placing clean filters into the sampler for 10 min without air flow. For seasonal analysis, the sampling period was divided as follows: summer (6 June to 28 August 2019, n=12), fall (4 September to 27 November 2019, n=13), winter (11 December 2019 to 26 February 2020, n=13), and spring (4 March to 6 May 2020, n=10). Meteorological data monitored by a Vantage Pro2 (Davis Instruments Corp., Hayward, CA, USA) weather station at the sampling site, including ambient temperature (T), relative humidity (RH), wind speed (WS) and wind direction (WD) were also obtained.

A 0.5 cm² punch from each filter was used to determine the concentration of organic carbon (OC) and elemental carbon (EC) in PM_2.5. The remaining portion of each filter was extracted by ultrasonication with 22 mL Milli-Q water (18.2 MΩ cm) for 40 min, followed by filtration through a 0.45 µm PTFE membrane filter (PALL Life Sciences, Ann Arbor, MI, USA). A suite of advanced instruments were employed to analyze the filtrate to characterize its chemical composition and optical properties, including a total organic carbon (TOC) analyzer for quantifying carbonaceous content, ion chromatography (IC) to measure water soluble ions, ultraviolet-visible (UV-Vis) spectroscopy to obtain absorbance spectra of water-soluble organic carbon (WSOC), excitation-emission matrix (EEM) fluorescence spectroscopy to assess the fluorescence characteristics of dissolved organic matter, and offline analysis using a high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) to investigate the detailed chemical composition of OA. Our measurement strategy aimed to capture the chemical composition of PM_2.5 as comprehensively as possible. The chemical species were broadly categorized into inorganic and organic constituents. Inorganic components were quantified using IC, while organic species were determined based on organic carbon measurements. Additionally, the optical properties of organic species were characterized using UV-Vis and EEM spectroscopy. Detailed descriptions of each measurement are provided below.

The total mass concentration of all species (WSIs + OM + EC) ranged from 2.0 to 41.8 µg m⁻³ during the study period, with a mean of 10.3 ± 7.4 µg m⁻³ (Fig. 2d). OM was the major contributor to aerosol mass concentration during the whole period with an average value of 37.7 %, followed by sulfate (21.3 %), nitrate (12.1 %), EC (1.1 %), and other inorganic ions, which together accounted for 27.8 % (including 7.5 % Na⁺, 7.6 % NH4+, 1.7 % K⁺, 6.6 % Ca²⁺, 0.8 % Mg²⁺, and 3.6 % Cl⁻). Seasonal Student's t-tests revealed that the mass concentrations during spring (14.0 µg m⁻³) and winter (12.5 µg m⁻³) were both significant higher than those in summer (7.1 µg m⁻³) and fall (8.0 µg m⁻³) (p<0.05) (Fig. 2c). These seasonal patterns were driven by increased transport of polluted air masses from the east in winter and prevalent mineral dust storms in spring. The natural mineral dust reached its peak in spring (1.05 µg m⁻³ of Ca²⁺) and its minimum in summer (0.29 µg m⁻³ of Ca²⁺) (Fig. 2c). The anthropogenic species (SO42- + NO3-) accounted for 33.2 % of the mass in spring and 32.8 % in winter. Among the secondary inorganic ions (sulfate, nitrate, and ammonium), sulfate was the most abundant, especially in summer, when its proportion reached 28.6 %, similar to observations conducted in July 2017 (Zhang et al., 2019). Sulfate formation during summer was mainly attributed to strong solar radiation, high humidity, and the heterogeneous reaction of SO₂ (Luo et al., 2022). In contrast, nitrate showed its minimum in summer (10.9 %) and its maximum in winter (15.5 %), which was mainly controlled by temperature-dependent partitioning. The average nitrate concentrations was 1.4 µg m⁻³ with 2.0 µg m⁻³ in spring, 0.8 µg m⁻³ in summer, 0.9 µg m⁻³ in fall, 1.9 µg m⁻³ in winter in this study, which are comparable to measurements at WLG in July 2017 (0.7 µg m⁻³) (Zhang et al., 2019) and at sites around the region, such as Qinghai Lake in the summer of 2010 (0.8 ± 0.5 µg m⁻³) (Li et al., 2013) and Menyuan in autumn 2013 (1.7 µg m⁻³) (Han et al., 2020). However, these concentrations are significantly higher than those recorded in the western Qilian Mountains, such as the summer 2012 observation at the Qilian Shan Station of Glaciology and Ecologic Environment (QSS) (0.6 µg m⁻³) (Xu et al., 2015).

Ion balance, represented by the ratio of cation equivalent concentration (CE, neq m⁻³) to anion equivalent concentration (AE, neq m⁻³), was used to assess potential missing ions or the acid-base properties of aerosols (Xu et al., 2014, 2015). The CE/AE ratio calculated in this study was 1.43 (Fig. 3a), suggesting the potential presence of acidic aerosols, although carbonate and bicarbonate ions were not measured in the IC analysis. Assuming 2 ⋅ [HCO3-] = [Ca²⁺], the estimated CE/AE is still 1.35. In addition, the ratio of [SO42- + NO3-] to [NH4+] was 1.90, indicating that there was an excess of sulfuric and nitric acids. The acidic property in the aerosol of our study can be further supported by a significant number of organic acids, such as oxalic acid (Fig. 3b). Oxalic acid is a product of atmospheric photochemical aging and is closely associated with sulfate and aerosol liquid water (Yang et al., 2009; Huang et al., 2019; Xu et al., 2020b; Boreddy et al., 2023). A significant correlation was found between oxalic acid peak area and sulfate during summer (Pearson's r = 0.61, P<0.05) (Fig. 3c).

Air mass backward trajectory analysis suggests that air mass origination varied from east to west seasonally, with the east mainly occurred during the summer transported with a shorter distance and the west during winter with a longer distance (Fig. 4). Specifically, the fraction of the air mass from the east was up to 50.5 % in spring and 66.0 % in summer and the mostly potential source areas for pollutants were predominantly associated with these air masses (Fig. 5). The less important source areas are also observed from the north and west, especially during the fall, when the climatic systems of summer monsoon and the westerlies interacted. In these directions, widely distributed mineral dust source areas and sparse urban cities are located. Overall, anthropogenic emissions located in the east of WLG emerge as the most significant source regions to the WLG.

The average absorption coefficient of WS-BrC at 365 nm (Abs₃₆₅) was 1.15 ± 0.97 M m⁻¹. The mean values for all seasons are shown in Table 1. The Abs₃₆₅ was much higher in spring and winter than in summer and fall (1.55 ± 1.30 M m⁻¹ in winter and 1.45 ± 0.54 M m⁻¹ in spring vs. 0.88 ± 0.70 M m⁻¹ in fall and 0.36 ± 0.21 M m⁻¹ in summer), which is consistent with the distribution of OM mass concentration. The results of the Student's t-test confirmed that the most pronounced seasonal difference in Abs₃₆₅ occurred between spring and summer (p < 0.01), followed by the pairs of summer-winter and fall-spring (p < 0.1). Note that the relationship analysis between WSOC and OC indeed exhibited a tight correlation (R2=0.81) and a relatively high ratio (0.63 ± 0.21), suggesting the representativeness of water-soluble fraction on bulk OM. The average absorption efficiency of WS-BrC at unit WSOC content (MAE) during the summer at 365nm (MAE₃₆₅, 0.40 ± 0.24 m² g⁻¹) is significantly lower than that of the other three seasons (0.92 ± 0.54 m² g⁻¹ in spring, 0.81 ± 0.46 m² g⁻¹ in fall and 0.97 ± 0.49 m² g⁻¹ in winter) (p<0.1) (Fig. 6a). A low MAE₃₆₅ value may result from either a reduced abundance of light-absorbing chromophores or photobleaching effects. However, we propose that photobleaching plays a predominant role, based on the observed negative correlation between MAE₃₆₅ and the oxidation state of OA (see Sect. 3.5). MAE₃₆₅ in summer is comparable to that at WLG (0.48 m² g⁻¹) in July 2017 (Xu et al., 2020a), Nam Co (0.38 m² g⁻¹) from 13 May to 1 July 2015 (Zhang et al., 2017a) and the regional background points of North China Plain (0.38 m² g⁻¹) in summer of 2017 (Luo et al., 2020). But the MAE₃₆₅ in spring of this study (0.92 ± 0.54 m² g⁻¹) is at a high level over the TP and even higher than the Qomolangma Station (QOMS) (0.81 m² g⁻¹) which is frequently impacted by biomass burning emission (Xu et al., 2020a).

AAE of light absorption spectrum is an important optical parameter to check the containing of BrC in aerosols. In the 300–400 nm range, a high AAE value indicates significant aerosol absorption of shortwave ultraviolet light, with a relatively higher contribution from BrC. This phenomenon is typically observed in cases from biomass burning emission, secondary organic aerosols (SOA), and anthropogenic pollutant emissions (Siemens et al., 2022; Tao et al., 2024). The AAE (300–400 nm) in this study ranges from 3.06 to 8.42, with an annual average of 5.42 ± 1.26 and peaking in summer (6.21 ± 1.50), followed by 5.48 ± 0.96 in winter, 5.19 ± 1.00 in fall, and 5.14 ± 1.46 in spring (Fig. 6a). The average annual AAE is comparable with the observation at Lulang (5.39 ± 1.22, 330–400 nm), in the southeastern part of the TP, during August 2014 to August 2015 (Li et al., 2016b) and Lhasa (5.38, 330–400 nm) during May 2013 to March 2014, a typical urban area on the TP (Li et al., 2016a). The summertime AAE is also similar to those at other stations on the TP, such as Nam Co (5.91 ± 2.14, 300–400 nm) from 13 May to 1 July 2015 (Zhang et al., 2017a) and WLG (5.96, 300–400 nm) from July 2017 (Xu et al., 2020a). The AAE value in winter is closed to that observed in the eastern Himalayas (5.5, 300–450 nm) during the 2019–2020 winter (Arun et al., 2024).

Figure 6b illustrates the comparison of optical properties of WSOA in the map space of AAE (300–400 nm) versus the logarithm of MAE₃₆₅ proposed by Saleh (2020). This map can be categorized into four classes as MAE₃₆₅ increase and AAE_300–400 decreases, which are associated with increased molecular sizes, decreased volatility, reduced solubility in water/organic solvents, and lower susceptibility to photobleaching. In our dataset, the majority of data points fall within the region corresponding to weakly absorbing brown carbon (W-BrC). This suggests that the OA at WLG were overall aged BrC, which is consistent with observations from other remote sites on the TP (Zhang et al., 2017a; Wu et al., 2020; Xu et al., 2022; Zhong et al., 2023). A limited number of data points, primarily from the spring, deviate from this pattern and fall closer to the region associated with moderately absorbing BrC (M-BrC). These outliers are likely influenced by episodic inputs of anthropogenic aerosols from surrounding populated areas.

PARAFAC analysis identified four components (C1–C4) in this study (Fig. 7a). The chemical properties of each component were determined based on the comparison with previous studies (Chen et al., 2016a, b, 2020; Yu et al., 2023; Zhong et al., 2023). C1 was determined as a highly oxidized humic-like component (HULIS-1) with Ex/Em maxima at 240/413 nm (Tang et al., 2024). C2 (Ex/Em = 225/375 nm) was classified as a less oxidized humic-like component (HULIS-2), typically associated with combustion-related source (Li et al., 2022; Afsana et al., 2023). Both C3 (Ex/Em = 280/358 nm) and C4 (Ex/Em = 225(270)/297 nm) were classified as protein-like organic matter (PLOM) (Wang et al., 2024). C3 was probably a fossil fuel-related substance (Wu et al., 2019), while C4 had a main peak and a secondary peak similar to the characteristic of tyrosine-like chromophore (Chen et al., 2016b, 2021b). HULIS compounds (C1 and C2) dominated the annual average contribution by 57.9 %, of which C1 accounted for 22.9 % and C2 accounted for 35.0 %. PLOM contributed an average of 42.1 %, with C4 accounting for 27.0 % and C3 being 15.1 % (Fig. 7b). C1 presented weak seasonal variation peaking in summer (23.54 %) corresponding to the highest intensity of photochemical oxidation. The average relative contribution of C2 was 37.0 %, 35.0 %, 33.6 % and 34.4 % in spring, summer, fall and winter, respectively. In contrast, C3 showed higher relative contributions in spring (17.8 %) and winter (17.0 %) than in summer (12.6 %) and fall (13.1 %), which may be related to frequent coal-burning emissions during heating period. Conversely, the contribution of C4 was significantly more pronounced during summer (28.9 %) and fall (30.5 %) than that in spring (23.4 %) and winter (25.1 %), corresponding to enhanced activities in agriculture emissions and ecological activities (Zheng et al., 2016; Zhang et al., 2020).

The fluorescence indices BIX and HIX serve as complementary tools for characterizing fluorescent OM. BIX is particularly used to assess the biological freshness of OM, whereas HIX reflects the degree of humification and chemical aging (Lee et al., 2013). By integrating these two indices, a more comprehensive understanding of the properties of OM can be achieved. An elevated degree of aging is associated with increased HIX values (Fan et al., 2020; Wu et al., 2021; Ma et al., 2022) and decreased BIX values (Wen et al., 2021). In this study, the annual average HIX and BIX values were 1.11 ± 0.18 and 1.29 ± 0.14, respectively, with seasonal variations of 1.04 ± 0.16 and 1.39 ± 0.24 in spring, 1.24 ± 0.11 and 1.26 ± 0.13 in summer, 1.13 ± 0.20 and 1.23 ± 0.09 in fall, and 1.02 ± 0.17 and 1.29 ± 0.09 in winter (Table 1). The spring samples exhibited the greatest variability, indicating their relatively fresh properties (Fig. 8b). In contrast, summer was characterized by the highest HIX, suggesting a high degree of aging and oxidation. The HIX of the autumn samples were at an intermediate level, while winter samples had the lowest HIX and moderate BIX values, indicating the lower degree of oxidation (Fig. 8a). Compared with previous studies, the fluorescence properties of aerosols in this study are more consistent with those in the northwestern China, and were less humified than those in the eastern China (Fig. 8) (Chen et al., 2021a; Zhang et al., 2021a; Zhong et al., 2023).

PMF decomposed the WSOA into two factors, i.e., a more oxidized oxygenated OA (MO-OOA) and a less oxidized oxygenated OA (LO-OOA) (Fig. 9a). Note that since our PMF analysis was conducted solely on the water-soluble fraction of OA, hydrophobic primary components may not be captured effectively. The mass spectra of these two OOAs in this study were consistent with those of online measurement at Nam Co Station in the TP during the summer (Xu et al., 2018). The average mass contribution of LO-OOA and MO-OOA were 47 % and 53 % (Fig. 9c), respectively. The mass contribution of MO-OOA across the four seasons (spring to winter) was 55.4 %, 54.9 %, 61.7 % and 42.0 %, respectively. The time series of LO-OOA correlated well with nitrate (R2=0.47) during winter and weak correlation with sulfate (R2=0.39), while MO-OOA correlated poorly with both species (Fig. 9b).

The triangle plot of m/z 44 (f44) versus m/z 43 (f43) and Van Krevelen diagram of elemental ratios are valuable tools for examining the ambient evolution of OA (Ng et al., 2010; Zhang et al., 2019; Chazeau et al., 2022). f44 is associated with highly oxidized oxygenated OA, while f43 corresponds to less oxidized OA. During atmospheric oxidation, OA generally evolves from a lower to a higher oxidation state, characterized by an increase in f44 and a decrease in f43, moving from the base to the apex of the triangular plot (Flores et al., 2014). In this study, most data points were located in the upper region of this triangular suggesting an overall high degree of oxidation across the dataset (Fig. 9e). Samples in winter were at the lower position and samples during summer shifted toward the higher position, presenting a distinct oxidation degree at different seasons. The Van Krevelen plot provides further insights into the chemical evolution of OA during atmospheric aging (Heald et al., 2010; Xu et al., 2018). The slope of the linear regression in the diagram reflects the predominant oxidation pathways. A slope between -1 to -0.5 is typically associated with functionalization processes such as the formation of carboxylic acids or alcohol/peroxide groups (Ng et al., 2011). In our dataset, the overall slope was -0.62 (Fig. 9f), indicating mixed contributions from both carboxylic acid and alcohol/peroxide formation pathways. This slope was higher than the value previously reposted for winter in Xining (-0.89) and summer at NamCo (-0.76) (Xu et al., 2018; Zhong et al., 2023). Seasonal slopes varied slightly, with spring and summer both at -0.58, fall at -0.60, and winter at -0.66, indicating the different OA oxidation pathways across seasons.

The light absorption characteristics of different WSOA factors, were evaluated by a multiple linear regression (MLR) model, which was applied to apportion their contributions to the light absorption at 365 nm (Abs₃₆₅) (Zhang et al., 2021b; Jiang et al., 2023). The MLR method can be expressed as Eq. (9). 9Absλ=f1×CMO-OOA+f2×CLO-OOA where fn is the corresponding fitting coefficients, which can also represent the mass absorption cross section (MAC) values of different organic components; CMO-OOAand CLO-OOA (µg m⁻³) are the mass concentration of the organic components; f×C is the absorption value of the organic component. The estimated MAC₃₆₅ were 0.41 m² g⁻¹ for MO-OOA and 0.45 m² g⁻¹ for LO-OOA (Fig. 6a). The MAC₃₆₅ value of LO-OOA was slightly higher than that of MO-OOA, which was related to the relatively weak photobleaching of LO-OOA. Compared to previous studies, MAC_365,MO-OOA in this study was lower than MAC_370,MO-OOA (0.60 m² g⁻¹) reported at the QOMS (Zhang et al., 2021b). MAC_365,LO-OOA was much lower than that observed at urban sites in northwestern China in winter 2019 (1.33 m² g⁻¹) (Zhong et al., 2023). The differences across sites could be attributed to variations in their chemical composition, sources, and also oxidation state.

During the atmospheric aging of BrC, changes in its optical properties can reflect concurrent alterations in its chemical characteristics (Alang and Aggarwal, 2024). In this study, we investigated the relationship between the MAE₃₆₅ and the elemental ratios of O/C and H/C across different seasons (Fig. 10). MAE₃₆₅ exhibited a positive correlation with O/C in spring (r=0.63; P<0.01) and a weak, statistically insignificant positive correlation in winter. In contrast, negative correlations were observed in summer and fall (r=0.29 and r=0.09). These seasonal relationships for MAE₃₆₅ and H/C showed an opposite pattern in each season (Fig. 10b). These results suggest that the light absorption capacity of BrC was enhanced during the oxidation process in spring and winter due to functionalization or oligomerization resulting in increased chromophore species, while further oxidation in summer and autumn leads to the fragmentation of large molecular weight compounds, leading to photobleaching and reduction in the light absorption capacity (Jiang et al., 2022).

In this study, cross-correlation analysis was performed on the chemical components of PM_2.5 (SO42-, NO3-, NH4+, K⁺, Cl⁻, LO-OOA, MO-OOA), oxidation degree (O/C and H/C), and the four PARAFAC-derived fluorescent components (C1–C4). As illustrated in Fig. S2 in the Supplement, C1, C2, and C3 exhibited positive correlations with the secondary species (SO42-, NO3-, NH4+, Cl⁻, LO-OOA, and MO-OOA). The strongest correlations were observed between C1 and these species, suggesting a secondary source origin for this chromophore. In contrast, C4 showed weak or no significant correlations with any of the chemical components, implying that C4 was not directly related to the oxidation processes of OA. As discussed in Sect. 3.3, C4 was related to agriculture emissions and ecological activities, rather than secondary atmospheric formation. To further investigate the optical evolution of WS-BrC during the atmospheric oxidation processing, the relationships of PARAFAC components and the WSOA components were integrated in EEM plot (Fig. 10c). C1 was associated with MO-OOA, whereas C2 and C3 were linked to LO-OOA, and C4 exhibited weak correlations with these two factors. This classification enables cross-validation of chemical and optical properties, providing additional insights into the formation pathways of chromophore components (Chen et al., 2016b; Zhong et al., 2023). Overall, the chemical transformation from less oxidized to highly oxidized OA through photochemical reactions can be extended to the process of BrC. Correspondingly, the optical evolution of BrC can serve as evidence of the oxidative state transition.