PM_2.5 samples were collected at the National Field Observation and Research Station (Shaanxi) for Ecological and Environment Changes and Integrated Management in the Guanzhong Plain Region, Xi'an, China (34°3^′39^′′ N, 108°20^′37^′′ E; 503 m a.s.l.). This station is positioned at the northern foothills of the Qinling Mountains (Fig. 1) and functions as a regional background site. It is located approximately 60 km southwest of Xi'an city, with no significant industrial, traffic, or other substantial anthropogenic pollution sources. The sampling campaign spanned from 1 January to 21 December 2022, and a total of 295 valid samples were obtained during the study. Daily sampling was conducted from 10:00 to 09:00 the next day, lasting for 23 h. Samples were collected using a medium-flow sampler (HC-1010, China Qingdao Company, China) equipped with quartz fiber filters (Whatman QMA, USA) with pre-baked (450 °C, 5 h) quartz fiber filters (Whatman, QMA, USA) at an airflow rate of 100 L min⁻¹. After sampling, each filter sample was promptly enveloped in aluminum foil, sealed, and frozen at -20 °C for subsequent laboratory analysis. To ensure quality control, field blank samples were collected both before and after sampling. This involved placing pretreated blank filters into the sampler, allowing them to stand for 10 min without activating the pump, and then sealing and storing them using the same procedure as the actual samples. The chemical analysis data for all samples were adjusted by deducting the average concentration measured from the two blank samples to eliminate potential systematic errors.

Daily average concentrations of PM₁₀, SO₂, NO₂, CO, and O₃-8 h (defined as the maximum 8 h average O₃ concentration) were obtained from the local government. Meteorological parameters, including ambient temperature (T), relative humidity (RH), wind speed (WS), and atmospheric pressure (AP), were acquired from the National Observation and Research Station for Regional Ecological Environment Change and Comprehensive Management in the Guanzhong Plain, Shaanxi Province, China.

Water-soluble inorganic ions (WSIIs), including SO42-, NO3-, NH4+, Cl⁻, Ca²⁺, K⁺, Mg²⁺, and Na⁺, were analyzed using ion chromatography (Metrohm-940, Switzerland). To enhance measurement accuracy, a fixed amount of lithium bromide was added to the eluent. For quality assurance, one sample out of every ten was randomly selected for re-analysis. All WSII concentrations were corrected using field blanks, which exhibited concentrations less than 10 % of those in ambient samples. The limits of detection (LODs) for SO42-, NO3-, NH4+, Cl⁻, Ca²⁺, K⁺, Mg²⁺, and Na⁺ were 0.027, 0.026, 0.0010, 0.0087, 0.0012, 0.0011, 0.0008, and 0.0005 µg m⁻³, respectively (Liu et al., 2025b). The mass concentrations of all detected WSIIs were significantly higher than their corresponding LODs.

As detailed in the preceding section, this study employed ammonium partitioning ratio (NHR) to quantify the transformation of gaseous ammonia into particulate ammonium (Xie et al., 2020; Zhang et al., 2021b). This ratio is defined as the molar concentration of particulate ammonium (n(NH4+), µmol m⁻³) relative to the total molar concentration of inorganic ammonium (n(NH3) + n(NH4+), µmol m⁻³), calculated using the following equation: 1NHR=n(NH4+)nNH4++n(NH3)

In this study, the ISORROPIA-II model was utilized to estimate aerosol thermodynamic parameters. The model was operated in “forward” mode, based on the assumption that aerosols were in a “metastable” state (Fountoukis and Nenes, 2007; Nenes et al., 1998), with inputs including concentrations of SO42-, NH_x, NO3-, Cl⁻, Na⁺, Ca²⁺, K⁺, Mg²⁺, RH and T. Aerosol pH was calculated using the following equation: 2pH=-lg1000γH+Hair+ALWC where γH+ is the hydronium ion activity coefficient (assumed =1), Hair+ (µg m⁻³) indicates the concentration of hydronium ions per unit volume of air, and ALWC (µg m⁻³) signifies the aerosol liquid water content. It should be noted that the model-calculated pH values are more reliable when the RH falls within the range of 30 % to 95 % (Guo et al., 2016).

Neglecting the influence of gaseous species, such as NH₃, HNO₃, and HCl, on aerosol acidity often results in an underestimation of acidity. The effect of these gases on aerosol pH can only be dismissed when their atmospheric concentrations are exceedingly low (Guo et al., 2015; Hennigan et al., 2015). In China, atmospheric ammonia concentrations are relatively high, and the study region is characterized by elevated ammonia levels; thus, NH₃ cannot be overlooked. In this study, We modified the method of Wei et al. (2023) by shifting the prediction target from NH₃ to NHR. Based on our earlier datasets (Wu et al., 2020), we performed multiple linear regression to estimate NHR, which was subsequently used to retrieve NH₃ levels. However, previous research has shown that the concentrations of gaseous HNO₃ and HCl in urban Xi'an remained below 1 µg m⁻³ throughout all seasons. Since the Qinling site serves as a background location, concentrations there are anticipated to be even lower. Given these negligibly low levels, we utilized the measured concentrations of particulate NO3- and Cl⁻ as inputs for TNO₃ and TCl, respectively (Nah et al., 2023).

In ammonia-rich regions, aerosol pH, particularly during winter, is primarily influenced by NH₃ levels (Tao and Murphy, 2019). However, the frequent absence of long-term and continuous NH₃ observational data impedes the analysis of long-term trends in PM_2.5 chemical properties. To fill this data gap and explore variations in aerosol pH in the research area, this study aimed to create a model for estimating the NHR based on easily accessible routine monitoring data to estimate NH₃ concentrations. Given the well-established thermodynamic and chemical relationships between NHR and factors such as NO3-, SO42-, and T (Fountoukis and Nenes, 2007), all parameters that serve as inputs to the thermodynamic model ISORROPIA-II, were included as candidate independent variables. Multiple linear regression (MLR) was performed using SPSS software with NHR as the dependent variable, and the model parameters were estimated using ordinary least squares.

The model was trained using a previous high-resolution (1 h) online PM_2.5 dataset from an urban site in Xi'an during 2016–2017 (Wu et al., 2020), which included major WSIIs concentrations in PM_2.5, NH₃ concentrations, and meteorological data (T and RH) from the same observation site. The data covered all four seasons: winter (n=605, from 21 December 2016 to 23 January 2017), spring (n=704, from 31 March to 30 April 2016), summer (n=591, from 8 July to 6 August 2017), and autumn (n=1032, from 1 October to 15 November 2017). Following model construction, its validity and accuracy were systematically assessed using the correlation coefficient, the F-test, and the T-test (Wei et al., 2023). For variables with p>0.05, they were retained to preserve model integrity, as they may indirectly influence NHR and were shown to have a negligible effect on pH verified by comparison (Table S1 in the Supplement). Despite the variations in ion concentrations between the training data and our study site, the thermodynamic mechanism that governs NHR remains consistent across different concentration levels. Furthermore, the ranges of the key input variables encompass the observed values at our site. Consequently, the model trained on urban data can be effectively utilized to predict NHR at the background site.

This study assessed the impact of various factors on aerosol pH through sensitivity tests. The variables analyzed included meteorological parameters (T, RH) and chemical components (SO42-, NH_x, NO3-, Cl⁻, Na⁺, Ca²⁺, K⁺, Mg²⁺) (Tao and Murphy, 2019). The specific sensitivity analysis adhered to the methodology outlined by Ding et al. (2019): Within the ISORROPIA-II model, the measured value of a target variable i was used sequentially as input alongside the average values of all other parameters. By observing the resulting trend in the output pH, the individual effect of variable i on aerosol pH could be assessed.

To quantitatively evaluate the influence of each factor on aerosol pH, two complementary approaches were utilized: the relative standard deviation (RSD) method and the ΔpH contribution decomposition method. The RSD, which is derived from the coefficient of variation of pH responses in sensitivity analysis (the ratio of standard deviation to mean), indicates the relative sensitivity of aerosol pH to typical fluctuations of a given factor; a higher RSD signifies a greater potential impact. However, the RSD method provides only a relative ranking and does not directly determine the absolute contribution of each factor to the observed pH. Consequently, this study emphasizes the quantitative analysis of ΔpH contributions quantitative analysis.

Assuming the aerosol pH estimated under scenario 1 (pH₁) differed from that under scenario 2 (pH₂), the pH difference (ΔpH = pH₂ - pH₁) was thus caused by the variations in the factors listed above. The quantitative analysis method adhered to the approach outlined by Zhou et al. (2022). To quantify the contribution of a single variable j within the ISORROPIA-II model, the value of factor j was altered from its Scenario 1 value to its Scenario 2 value, while all other parameters remained constant. The resulting change in pH was recorded as ΔpH_j, which represents the independent contribution of that specific factor to the total ΔpH. In this study, the aforementioned method was applied to consecutive seasonal pairs in the following order: spring values were substituted into the winter scenario, summer values into spring, autumn values into summer, and winter values into autumn (Table S2). For each pairwise substitution, Scenario 2 denotes the season being introduced, while Scenario 1 refers to the base season. Consequently, the calculated ΔpH quantifies the impact of each factor's seasonal shift on aerosol pH. The remaining change in pH, not accounted for by this procedure, is attributed to ΔpH_others, reflecting the combined effects of synergistic variations among the factors.

As a crucial precursor of secondary particulate matter, NH₃ influences aerosol acidity through the multiphase buffering of its NH₃/NH4+ conjugate pair (Liu et al., 2019; Saraswati et al., 2019; Zheng et al., 2020). This buffering effect is especially pronounced in the ammonia-rich Guanzhong Plain (Zhang et al., 2021c; Zheng et al., 2023). However, the lack of online monitoring data for gaseous NH₃ at the sampling site limits the accuracy of pH estimation using thermodynamic models. Therefore, it is important to establish a reliable method for estimating NH₃. In this study, we adapted the methodology of Wei et al. (2023) by replacing the prediction target from NH₃ to the thermodynamic state parameter NHR, and estimated the NHR using multiple linear regression analysis based on our previous short-term scale datasets (Wu et al., 2020). This is because the direct estimation of NH₃ is prone to interference from meteorological factors and the gas-particle distribution of acidic precursors, resulting in unstable outcomes in ammonia estimation. Consequently, this study opts to estimate NH₃ through NHR and assesses the method's accuracy (Meng et al., 2018; Wei et al., 2023).

Tables 2 and S4 summarize the total results of multiple linear regression analysis in each season. The F-test and t-test results reveal significant correlations for all seasons (R>0.82, P<0.01), confirming the efficacy of the established statistical regression equations. Subsequently, a Pearson correlation analysis was performed to elucidate the statistical associations between NHR and diverse environmental factors. As shown in Fig. 3 and Table S5, NHR exhibited significant correlations (P<0.01) with all selected environmental factors in both winter and spring. Among these, the highest correlation coefficient was observed between NO3- and SO42- (R=0.96, P<0.01). Across all four seasons, NHR showed good correlations with either NO3- or SO42-, but with distinct seasonal variations. In winter and spring, NHR displayed the highest Pearson correlation coefficients with NO3- (R=0.86, P<0.01 in winter; R=0.74, P<0.01 in spring). Conversely, in summer and autumn, NHR correlated most significantly with SO42- (R=0.75, P<0.01 in summer; R=0.79, P<0.01 in autumn). These seasonal distinctions primarily arise from fluctuations in the thermodynamic stability of different ammonium salts and the seasonal alterations in their formation pathways (Tao et al., 2025; Zhang et al., 2021c).

This study assessed the reliability of the MLR model estimates through two validation levels. First, the NHR estimated by the MLR model for 2016–2017 was compared to the observed NHR (Fig. 4). The results revealed a significant positive correlation across all seasons (with the intercept constrained to zero, R2>0.94, P<0.01), indicating that the MLR model predicts NHR with high accuracy. Second, to further validate the pH results at our site, we compared the NHR output from the ISORROPIA-II model with the input NHR (Fig. S3). The results again demonstrated a linear correlation (with the intercept constrained to zero, R2>0.83, P<0.01), thereby supporting the reliability of subsequent pH simulations. However, a disparity exists between the predicted and observed NHR in Figs. 4 and S3. To evaluate its impact on aerosol pH, we calculated the pH changes associated with both positive and negative NHR deviations (Table S1). The overall pH change resulting from these deviations is less than 0.2 units, which does not affect the primary conclusions. Furthermore, the NH₃ concentrations derived from NHR at our site remain below 60 µg m⁻³, well within the range of the training dataset from urban site. However, the seasonal NH₃ pattern at our site (2022: winter > summer > spring > autumn) differs from that in the training dataset (2016: summer > spring > autumn > winter). This discrepancy can be attributed to the significant reduction in SO₂ emissions resulting from air pollution control. In 2016, the high concentration of SO₂ emitted from coal combustion reacts rapidly with NH₃ to form (NH₄)₂SO₄, and this process consumes large amounts of gaseous NH₃ in winter (Wang et al., 2016; Liu et al., 2023). Due to interannual NH₃ variability, we employed the thermodynamically stable NHR for model training and validation, back-calculated NH₃ from the estimated NHR and measured NH4+, and then used these values in ISORROPIA-II to simulate aerosol pH.

Figure 5 shows the monthly averages of aerosol pH, aerosol liquid water content (ALWC), and Hair+, as calculated by the ISORROPIA-II model during the 2022 sampling period. The annual average aerosol pH at the Qinling observation site was 3.8 ± 1.0. The seasonal pH averages ranked as follows: winter (4.8 ± 0.5) > spring (3.6 ± 0.5) > autumn (3.3 ± 0.7) > summer (2.7 ± 0.7). The average ALWC concentration was 60.3 µg m⁻³, exhibiting the pattern: winter (80.4 ± 112.9 µg m⁻³) > autumn (76.2 ± 121.0 µg m⁻³) > spring (55.7 ± 132.2 µg m⁻³) > summer (10.0 ± 25.1 µg m⁻³). Overall, aerosol pH was highest in winter and lowest in summer, with spring and autumn values intermediate. This trend closely aligns with the seasonal evolution of PM_2.5 chemical composition and meteorological conditions, reflecting the seasonal pH variation pattern observed in Beijing (Ding et al., 2019). Winter aerosol pH was similar to those reported in Xi'an previously and slightly higher than those recorded during the COVID-19 period (Liu et al., 2025b). A comparison of aerosol acidity levels across various regions in China is detailed in Table S6, revealing substantial regional variability. In contrast to southeastern coastal cities such as Shanghai and Xiamen, typical inland cities like Xi'an and Beijing generally display higher aerosol pH (Fu et al., 2022; Liu et al., 2025b; Xu et al., 2025).

To further explore the seasonal characteristics in PM_2.5 acidity/alkalinity under different pollution levels, this study classified PM_2.5 mass concentrations into two categories based on ambient air quality standards of China: clean days (<75 µg m⁻³) and polluted days (>75 µg m⁻³). A systematic comparison of percentage change in major WSIIs, aerosol pH and ALWC variations corresponding to these pollution levels across different seasons was performed (Table S7, Fig. S4). Under clean conditions, aerosol pH ranged from 2 to 5, whereas under polluted and heavily polluted conditions, it was primarily concentrated between 3 and 6. Independent t-tests confirmed significantly higher pH on polluted days for winter, spring, and autumn (p<0.01). In summer, only two polluted days were recorded, so the statistical results for summer (Table S7) should be interpreted with caution due to the very limited sample size a reliable t-test could not be performed. As air quality declined and PM_2.5 concentrations increased from clean to polluted days, the concentrations of aerosol components and pH values exhibited an upward trend across all seasons. However, the behaviors of RH, ALWC, and Hair+ displayed seasonal discrepancies. In spring and summer, RH, ALWC, and Hair+ decreased as PM_2.5 concentrations increased. Conversely, during autumn and winter, these parameters increased alongside rising PM_2.5 concentrations. These findings reveal fundamental differences in the dominant chemical mechanisms of PM_2.5 pollution formation across seasons.

During spring and summer, pollution formation is mainly controlled by gas-phase photochemical oxidation (Chen et al., 2025; Li et al., 2021). The enhanced solar radiation and high temperatures during these seasons collectively create a highly oxidizing atmospheric environment. While promoting the transformation of precursors, these conditions also drive the decomposition of semi-volatile NH₄NO₃, leading to the release of gaseous HNO₃ and NH₃ (Guo et al., 2018; Tao et al., 2025). Consequently, despite an increase in PM_2.5 concentrations, both ALWC and Hair+ concentrations exhibit a decreasing or stabilizing trend due to the reduction of hygroscopic nitrate, while RH remains relatively low because of higher T. In contrast, autumn and winter are characterized by pollution processes driven by heterogeneous and aqueous-phase chemical reactions. The cold temperatures and stagnant weather conditions lead to high RH, promoting the aqueous-phase oxidation of pollutants and inhibiting the evaporation of NH₄NO₃ (Wang et al., 2016; Zhang et al., 2021c). This results in the significant formation of highly hygroscopic NH₄NO₃ and (NH₄)₂SO₄, which triggers an increase in ALWC. This increase further intensifies heterogeneous reactions and concentrates secondary acidic species such as H₂SO₄ and HNO₃, leading to a concurrent rise in Hair+ concentration. Consequently, a complex pollution scenario emerges, characterized by a simultaneous increase in PM_2.5 ALWC, and Hair+ concentrations.

Sensitivity analysis of aerosol pH and ALWC in response to meteorological conditions and chemical composition is a crucial step for understanding and predicting atmospheric environmental changes. This analysis not only clarifies the nonlinear processes involved in secondary aerosol formation but also assesses the impact of variations in anthropogenic emissions on atmospheric chemical systems (Nah et al., 2023; Zhou et al., 2022). In this study, sensitivity analyses were performed for each season, incorporating both particulate-phase compositions and meteorological factors. The effects of SO42-, NO3-, NH_x (gas NH₃ + NH4+), Cl⁻, Ca²⁺, K⁺, Mg²⁺, RH, and T on aerosol pH, ALWC, and Hair+ were evaluated, as illustrated in Fig. 6. The primary driving factors influencing aerosol pH variation across all seasons were NH_x, SO42-, and T, while Ca²⁺ in spring and RH in summer also emerged as important influencing factors. This pattern is consistent with the driving factors identified in the Beijing area (Ding et al., 2019). Regarding ALWC, RH emerged as the most significant factor, followed by NO3- and SO42-.

Sulfate as a key component of secondary inorganic aerosols, is one of the most significant acidic sources in atmospheric fine particulate matter. Sulfuric acid, generated through heterogeneous oxidation processes, is a strong acid that directly increases the concentration of Hair+ in the particulate phase and reduces the pH (Shah et al., 2018; Zhang et al., 2021c). As illustrated in Fig. 6, when the levels of alkaline substances such as NH₃ are maintained constant, aerosol pH significantly declines with increasing SO42- concentration. The concurrent increase in ALWC and Hair+ confirms the sulfate-driven acidification mechanism: SO₂ is converted to H₂SO₄ via heterogeneous oxidation, and the dissociation of H₂SO₄ releases H⁺, leading to increase Hair+ and decrease pH. (Ding et al., 2019). The increase in ALWC and Hair+ is particularly pronounced during summer and autumn. This sustained rise continuously depletes alkaline buffering substances in the particulate phase, resulting in increased acidity (Guo et al., 2018). Conversely, while nitrate similarly elevates Hair+ and ALWC, its overall impact on pH is generally weaker than that of sulfate due to its high volatility (Ding et al., 2019). Except during summer when the pH remained relatively stable, pH values decreased with rising NO3- concentrations in all other seasons.

Ammonia (NH₃), the primary alkaline gas, plays a crucial role in buffering aerosol pH. In ammonia-rich environments, NH₃ effectively counteracts acidic components, stabilizing pH and resulting in a nonlinear pH response to precursor concentrations (Cui et al., 2025; Karydis et al., 2021; Niu et al., 2016). Simultaneously, the dissolution of NH₃ into the particle phase to form NH4+ markedly enhances the hygroscopicity of particles, causing an increase in ALWC as NH_x concentrations rise (Jia et al., 2020; Liu et al., 2019). Studies indicate that NH₃ first reacts with H₂SO₄ and subsequently with HNO₃ to form NH₄NO₃. After most nitrate is converted to NH₄NO₃, it becomes difficult to dissolve additional NH₃ into aerosol droplets (Ding et al., 2019; Xu et al., 2025). The results of this study (Fig. 6) show that the response of pH to NH_x is non-linear. When the NH_x mass concentration is within the range of 3–10 µg m⁻³, aerosol pH increases significantly, Hair+ decreases sharply, and ALWC increases with rising NH_x concentration. Nevertheless, when the NH_x concentration surpasses a specific threshold, the trends in pH and Hair+ decelerate and subsequently remain relatively stable. It should be noted that the results may be less accurate when NH_x concentration is below 3 µg m⁻³, as high temperature and intense solar radiation promote the thermal decomposition of semi-volatile ammonium salts. The seasonal variations observed in this study further substantiate the regulatory role of atmospheric ammonia.

Ca²⁺ is an important crustal element in mineral dust, displaying distinct regional chemical characteristics and effects on aerosol acidity. According to the ISORROPIA-II model, Ca²⁺ typically combines with SO42- to form the less soluble CaSO₄ solid phase, reducing its solubility and diminishing its effectiveness for neutralizing atmospheric acidity (Fountoukis and Nenes, 2007; Wang et al., 2016). In the Guanzhong Plain, the regulatory impact of Ca²⁺ on aerosol pH is particularly crucial, especially due to the region's proximity to the Loess Plateau and its semi-arid climate. Dust transport from the northwest could shape the alkaline atmospheric conditions in this area. This phenomenon results in neutral aerosol pH in the free troposphere and contributes to the alkalinity of cloud water at high-altitude sites (Liu et al., 2024; Shen et al., 2024). Elevated concentrations of Ca²⁺ can increase aerosol pH by decreasing both ALWC and Hair+, with the most pronounced impact observed during the frequent dust events of spring. These events transport significant amounts of alkaline mineral dust, enhancing acid neutralization, consistent with studies in northern China (Shi et al., 2019; Tan et al., 2018). The relative standard deviation (RSD) can also be used to quantify the sensitivity of aerosol pH to a specific factor, with a higher RSD indicating a greater influence. This study revealed that RSD for Ca²⁺ concentration variation in spring reached 8.8 % (Table S8), notably higher than the corresponding RSD of 7.5 % reported for the North China Plain during the same season (Ding et al., 2019). Conversely, in southern coastal urban areas like Xiamen and Guangzhou, which lack robust dust origins, the impact of Ca²⁺ on aerosol pH is typically negligible (Jia et al., 2020; Xu et al., 2025). This finding directly confirms the greater influence of dust sources in the Guanzhong Plain.

In addition to the particulate chemical composition, meteorological conditions, specifically RH and T, significantly influence aerosol acidity. This study revealed that during winter, spring, and autumn, aerosol pH displayed a non-linear response to increasing RH, characterized by an initial decrease followed by a subsequent increase. The inflection point typically occurred within the RH range of 60 %–85 %. Beyond this critical threshold, ALWC entered a phase of exponential growth. This phenomenon is probably driven by the deliquescence of the predominant hygroscopic components within the aerosol, primarily NH₄NO₃ and (NH₄)₂SO₄, which have deliquescence relative humidities of approximately 62 % and 80 % (Tang, 1996). Notably, the inflection point exhibits clear seasonal variations, which can be attributed to seasonal shifts in aerosol chemical composition (Table 1). Notably, the inflection point varies seasonally maybe due to the varying proportions of NO3- and SO42- in SNA: high NO3- in autumn and winter shifts the inflection point toward lower RH (∼ 62 %), while the higher SO42- contribution in spring shifts it toward higher RH (∼ 80 %). Given that actual atmospheric particles are mixtures of various components, their hygroscopic and phase-state characteristics are more complicated (Su et al., 2022; Zheng et al., 2020). In the initial stage, an increase in RH results in a higher ALWC, which facilitates the heterogeneous uptake and aqueous-phase oxidation of acidic precursors, resulting in the formation of strong acids. During this phase, the acid-forming effect dominates the pH change (Du et al., 2020; Zhang et al., 2021c). Once RH exceeds a critical threshold, ALWC exhibits an exponential growth trend. At this point, the intense dilution effect of RH on H⁺ begins to surpass the ongoing acid-forming effect (Karydis et al., 2021). Furthermore, this hygroscopic growth is self-reinforcing, i.e. the increased ALWC provides a more extensive aqueous medium for heterogeneous reactions of gaseous precursors, leading to further formation of hygroscopic salts and further promoting ALWC growth, and thus establishing a positive feedback loop (Su et al., 2022). Concurrently, under high RH conditions, thermodynamic factors favor the partitioning of semi-volatile ammonium salts into the particle phase. This neutralization process further depletes H⁺ (Tao et al., 2025). These combined effects ultimately lead to the observed recovery in pH. Ambient temperature can influence aerosol pH by affecting gas-particle partitioning (Ding et al., 2019). The evaporation of semi-volatile ammonium salts plays a crucial role in aerosol acidification by transferring H⁺ to the particle phase, consequently decreasing pH (Guo et al., 2018). Elevated ambient temperatures intensify this acidification mechanism synergistically. Thermodynamically, elevated temperatures drive the volatilization process. Concurrently, they reduce ALWC to concentrate H⁺ and further enhance acidity.

To quantitatively analyze aerosol pH, the contributions of various influencing factors to aerosol pH and ALWC were further quantified (Figs. 7, S5; Table S2). According to the multiphase buffer theory (Zheng et al., 2020), T significantly drives the seasonal variation of aerosol pH by influencing pKa*. The results presented in Fig. 7 confirm this assertion, indicating that T plays a dominant role (22.3 %–31.6 %) in driving the seasonal variation of aerosol pH. T favored enhanced aerosol acidity in winter and spring, while the opposite trend was observed in summer and autumn. Moreover, NH_x emerged as a key factor influencing aerosol pH during winter and autumn, contributing between 22.9 % and 44.9 %. When the higher winter NH_x concentration (pH₁) was replaced with the lower concentration observed in spring (pH₂), while maintaining other winter parameters constant, the simulated ΔpH <0 (pH₂ - pH₁), indicating increased acidity. In the winter, the widespread use of coal for heating leads to significant emissions of acidic gases (Wang et al., 2016). The decrease in NH₃ concentrations may exacerbate acidification through two mechanisms. On one hand, lower NH₃ weakens the buffering capacity of the NH₃/NH4+ conjugate pair, which is the dominant alkaline buffer system for aerosol pH (Zheng et al., 2020). Less NH₃ is thus available to neutralize H⁺ from SO₂ and NO_x oxidation, making the system more vulnerable to acidification. On the other hand, reduced NH₃ disturbs the gas-particle equilibrium between particulate ammonium nitrate and gaseous ammonia along with nitric acid. According to Le Chatelier's principle, a decrease in gas-phase NH₃ drives this equilibrium toward dissociation, promoting the decomposition of particulate NH₄NO₃ and releasing additional gaseous nitric acid (Guo et al., 2018). These two processes together contribute to a decline in system pH. In contrast, the opposite trend is observed in the autumn. The primary influencing factors in spring were T at 31.9 % and NVCs at 24.7 %, while in summer, they were T at 33.8 % and RH at 17.8 %. These results suggest that NVCs, represented by Ca²⁺, are rich in alkaline mineral components and can enhance acid-neutralizing capacity, making them important drivers of aerosol pH variations in spring. In summer, RH alters aerosol acidity by affecting the deliquescence of hygroscopic particles, thereby serving as a key driver of aerosol pH variations. This observation is consistent with findings from studies conducted in other northern Chinese cities (Ding et al., 2019). Moreover, the combined effects of factor interactions and model nonlinearity, represented by the residual contribution (ΔpH_others) are also significant, particularly in autumn, where they contributed 26.6 % to aerosol pH variation.

Throughout the observation period, chemical components contributed an average of 47.7 % annually to pH, while meteorological factors contributed 35.7 %, highlighting the greater impact of chemical components over meteorological factors. This characteristic significantly differs from that observed in southeastern coastal regions of China. For example, studies conducted in Xiamen and Guangzhou have demonstrated that the contribution of meteorological factors frequently surpasses that of chemical components. This difference arises from regional characteristics. In coastal areas, meteorological factors, including RH and sea-land breezes, exhibit significant variability, while the chemical composition remains relatively stable, rendering meteorological conditions predominant. Conversely, in inland semi-arid regions, chemical components such as NH_x and dust-derived Ca²⁺ exhibit pronounced seasonal fluctuations, resulting in a greater dominance of chemical factors (Jia et al., 2020; Xu et al., 2025). Seasonally, chemical components were more influential in winter (58.7 %), spring (42.7 %), and autumn (49.2 %). Conversely, meteorological factors displayed their most significant impact in summer, with a contribution rate reaching 51.6 %, thus serving as the primary driver of pH variation during this season.