2.3.1 Online measurements by TOF-ACSM and MARGA

The non-refractory (NR) chemical compositions of PM_1.0 and PM_2.5 were measured continuously during the RITA-2021 campaign with a time resolution of 5 min, including ammonium (NH${}_{\mathrm{4}}^{+}$), nitrate (NO${}_{\mathrm{3}}^{-}$), sulfate (SO${}_{\mathrm{4}}^{\mathrm{2}-}$), chloride (Cl⁻), and organics (organic aerosol, OA), using two TOF-ACSMs (Aerodyne Research Inc., Billerica, MA) (Fröhlich et al., 2013), both equipped with a CV and PM_2.5 aerodynamic lens (Xu et al., 2017). The two TOF-ACSMs were installed side by side in a trailer, which was around 200 m away from the above-mentioned main measurement site. Teflon-coated aluminum cyclones (URG 2000-30ED) were installed at the head of the inlet with a downward entry direction to avoid external effects such as rain. Flow rates of 2.3 and 5.0 L min⁻¹ were applied to select PM_2.5 and PM_1.0, respectively. Then, a multi-tube Nafion dryer (Perma Pure, New Jersey) was used to dry the particles. It should be emphasized that the size selection occurred at ambient conditions; thus the upper limit of the dry particle size depends on humidity. The working principle of the TOF-ACSM is based on the Aerodyne aerosol mass spectrometer (AMS) and can be briefly described as follows: the particles are focused and drawn into the instrument through an aerodynamic lens, and then the non-refractory constituents are evaporated rapidly by the capture vaporizer (T=600 °C) and subsequently ionized by electron impact. The ions are identified by their mass-to-charge ratio in the time-of-flight mass spectrometer. In the end, the electrical signal is converted to a digital signal by the signal detector and recorded (Fröhlich et al., 2013). Several calibrations need to be performed regularly to ensure the accuracy of the instruments, including the lens calibration, flow rate calibration, and tuning of the heater bias (HB) voltage, as well as the ionization efficiency (IE) and the relative ionization efficiency (RIE) calibrations. The standard procedure of the calibration details can be found in previous publications (Fröhlich et al., 2013; Canagaratna et al., 2007). The IE calibration and RIE calibration were performed before the RITA campaign, and the parameters used in this paper are summarized in Table 1. The data analysis was produced by Tofware (v3.2.4, Tofwerk AG, Thun, Switzerland) based on Igor Pro 8.

Table 1The setup details for two of the TOF-ACSMs and the corresponding IE and RIE calibration values for each species used in this study.

Download Print Version | Download XLSX

The Monitor for AeRosols and Gases in ambient Air (MARGA 2060, Metrohm Applikon B.V., the Netherlands) was used during the September part of the campaign to measure the water-soluble inorganic components based on ion chromatography (IC), including hydrochloric acid (HCl), nitric acid (HNO₃), nitrous acid (HONO), sulfur dioxide (SO₂), and ammonia (NH₃) in the gas phase and chloride (Cl⁻), nitrate (NO${}_{\mathrm{3}}^{-}$), sulfate (SO${}_{\mathrm{4}}^{\mathrm{2}-}$), ammonium (NH${}_{\mathrm{4}}^{+}$), potassium (K⁺), calcium (Ca²⁺), and magnesium (Mg²⁺) in the aerosol phase. A “MARGA sizer” introduced by ten Brink (2007, 2009) was used to control the size of the particles (e.g., PM_1.0, PM_2.5, or PM₁₀) entering the instrument. We applied the PM_1.0 sizer in the first stage (from 5 to 30 September 2021) and PM_2.5 sizer in the later stage (from 3 to 16 October 2021) of the campaign. The ambient air was drawn into the instrument at a constant flow rate of 16.7 L min⁻¹ through a short (0.2 m) length of Teflon tubing with an outer diameter of 25.4 mm, via a vacuum pump. Then, the water-soluble gases were absorbed by a wet rotating denuder (WRD) device (Wyers et al., 1993; Keuken et al., 1988), and the water-soluble aerosols were extracted in a steam-jet aerosol collector (SJAC) (Khlystov et al., 1995; Slanina et al., 2001). Eventually, the liquid of the samples was collected continuously in separate syringes and then analyzed by IC at 1 h resolution. Rumsey and Walker (2016) and Rumsey et al. (2014) provided the operational, calibration, and data analysis procedures in detail.

2.3.2 Offline filter measurements and analysis

From midnight to midnight, 24 h PM_1.0 and PM_2.5 filters were collected simultaneously, according to the reference method described in the European Standards (EN12341: 1998 and EN14907: 2005). The SEQ47/50 (Leckel GmbH, Germany) instrument with a sequential low-volume system (LVS) of 2.3 m³ h⁻¹ was used for the sampling. Two polytetrafluoroethylene (PTFE) filters (diameter 47 mm, pore size 3 µm, Millipore) and two quartz fiber filters (diameter 47 mm, Pallflex) were placed in the samplers for a paired measurement of PM_1.0 and PM_2.5. The quartz filters were pre-baked at 550 °C over 6 h before use. All filter samples were collected and stored at $-\mathrm{20}\pm \mathrm{0.5}$ °C and cooled by ice packs during transportation. The gravimetric mass of the PTFE filters was obtained by triple weighing before and after sampling. The weighing was performed under a condition of a temperature of 20.0±0.5 °C and relative humidity of 50±2 %. Detailed information about the logistic and operational (QA $/$ QC, weighing) procedures, as well as the data acquisition was described in Schaap et al. (2010).

The PM_1.0 and PM_2.5 quartz fiber filters were used to perform the ion analysis by ion chromatography, including the three inorganic anions (NO${}_{\mathrm{3}}^{-}$, Cl⁻, SO${}_{\mathrm{4}}^{\mathrm{2}-}$) and the five cations (Na⁺, K⁺, Mg²⁺, Ca²⁺, NH${}_{\mathrm{4}}^{+}$). Aliquots of the filter samples (cations: 3.28 cm²; anions: 3.0 cm²) were extracted by the 2.0 mL of the 30 mM methane sulfonic acid (MSA; cations) or 2.0 mL of extra pure water (anions) for 40 min under ultrasonic agitation. The determination of the concentrations was performed by the ICS-1100 and AQUION instruments (both Thermo Scientific) combined with an autosampler AS-DV and ion exchange columns (cations – CS16, anions – AS22). The description of the equipment used can be found in Samek et al. (2020). The organic carbon (OC) and elemental carbon (EC) were analyzed by a Sunset thermal–optical analyzer (TOA; Sunset Laboratory Inc.) on the quartz fiber filters. The EUSAAR2 protocol (Cavalli et al., 2010) was applied to distinguish the OC and EC. The details of the operation procedure can be found in Yao et al. (2022).

3.1.1 Comparison of online and offline measurements for PM_1.0 and PM_2.5

Figure 1 shows the comparison of each component between the online and offline measurements for PM_1.0 (green dots) and PM_2.5 (orange triangles), including a linear least-squares regression line. The uncertainties of the slope correspond to the standard error. The daily fluctuations in the online measurements are shown in Fig. S7 in the Supplement. We illustrate this by presenting the standard deviation of the daily measurements, which are taken at 10 min intervals. These variations are marked as error bars for each individual day. Over the intensive measurement period, the daily average NO₃ mass concentrations measured by the TOF-ACSM-CV and by filters showed a high correlation with R²= 0.98 for PM_1.0 and R²= 0.97 for PM_2.5, and the corresponding slopes are 0.94 ± 0.09 and 0.88 ± 0.10, respectively. The results showed that NO₃ concentrations on the filters were slightly higher than TOF-ACSM-CV measurements. Paired t tests were performed to investigate the significance of the difference between the online and offline measurements, and the results are shown in Table S1 and S2 in the Supplement. It shows a significant difference between the ACSM-measured NO₃ and filter-measured NO₃ (p values are $\mathrm{2.38}\times {\mathrm{10}}^{-\mathrm{5}}$ for PM_1.0 and $\mathrm{5.08}\times {\mathrm{10}}^{-\mathrm{7}}$ for PM_2.5). Since the filter concentrations were higher, the difference cannot be explained by the evaporation of ammonium nitrate collected on the filter, which is a well-known sampling artifact (Malaguti et al., 2015; Kuokka et al., 2007; Chow et al., 2008; Pakkanen and Hillamo, 2002). Previous studies have shown that the loss grows with an increase in temperature and a decrease in humidity and that it can exceed 80 % up to complete evaporation when the temperature exceeds 25 °C (Schaap et al., 2004; Allan et al., 2003; Pandolfi et al., 2014). During the measurement in this study, the RH was 81.16 ± 14.17 %, and temperature was 15.94 ± 4.20 °C, which should largely prevent this evaporation loss. Consequently, we observe slightly higher concentrations on the filter samples. Likely reasons for this difference are that (i) the higher offline concentration of nitrate may be caused by the absorption of gas-phase nitric acid (HNO₃) on the filter (Chow, 1995). Bhowmik et al. (2022) also observed higher nitrate concentrations on filter samples with an even lower slope of 0.49 between the online AMS and offline filter NO₃ measurements. (ii) For ACSM measurements, the absolute concentration of the nitrate is highly dependent on the IE calibration, which needs to be performed carefully and regularly. The calibration parameters used in this study are listed in Table 1. If they are slightly biased, the ACSM concentration could be too low. However, the differences in filter and ACSM NO₃ are in general less than 10 %, which are much better than the previous ±30 % accuracy given in the manufactory for the ACSM with a SV and PM_1.0 lens.

Figure 1The linear regression fitting correlations between the online (ACSM and MAAP) and offline (filters) daily average mass concentrations of various chemical components. PM_1.0 is indicated in green and PM_2.5 in orange. The shaded area represents the 95 % confidential interval of the best fit line.

Download

For sulfate, the online and offline measurements also showed a high correlation, though it was lower than for ammonium nitrate. The slope and coefficient of determination are 0.90 ± 0.16 and R² is 0.93 for sulfate PM_1.0, and the slope is nearly 1 (0.99 ± 0.24) and the R² is 0.87 for sulfate PM_2.5. The relatively lower R² is potentially due to the low sulfate mass concentration (0.67 and 0.84 µg cm⁻³ on average for PM_1.0 and PM_2.5) during the measurements. Similar to nitrate, the ACSM sulfate measurements are influenced by the IE and RIE calibrations. Apart from that, higher offline values of the sulfate may also be caused by some refractory sulfates such as potassium sulfate, calcium sulfate, and sodium sulfate, which cannot be detected by TOF-ACSM (Poulain et al., 2020). Or it can also be due to the positive sampling artifacts, for example, the absorption of SO₂ by alkaline particles in the filter membrane or by the reaction of gas-phase ammonia with sulfate aerosols to form ammonium sulfate or ammonium bisulfate (Nicolás et al., 2009; Nie et al., 2010). This is less likely to occur in the Netherlands as sulfate is usually completely neutralized by excess ammonia already in the ambient atmosphere.

For ammonium, the coefficients of determination were R²= 0.98 in PM_1.0 and R²= 0.94 in PM_2.5 with slopes of 1.09 ± 0.10 and 0.96 ± 0.15, respectively. As the ammonium measured by the ACSM mainly corresponds to ammonium nitrate and ammonium sulfate, the small deviation of the online and offline data is reasonable. However, it is worth noting that the ammonia concentrations in Europe as a whole are usually sufficient to neutralize nitric and sulfuric acid (Wichink Kruit et al., 2017). In particular, an excess of ammonium (ammonium concentrations higher than those explained by the formation of inorganic ammonium salts) has been observed a lot in the Netherlands in past reports (Schlag et al., 2017). Tables S4 and S5 show the molar mass concentration of cation (NH₄) and anions (NO₃ and SO₄) from the filter samples and ACSM measurements. Anions are observed to be 7 % higher than cation in PM_1.0 filter samples, indicating a light underestimation of NH₄ in filter PM_1.0. But on the whole, the average differences between the ACSM and filter samples are less than ±10 % for all inorganic chemical species, showing a good accuracy of the ACSM with the CV and PM_2.5 lens in the field measurements.

Regarding the measurement of the organic aerosol (OA) fraction, the ACSM measures OA, under the assumption that all mass, which cannot be explained by known inorganic components, must be organic (Allan et al., 2004). Thus, the quantification of the OA concentration is determined by how to interpret and assign fragmentation signals. On the other hand, the offline measurement of the organics is normally done by thermal–optical analysis, which usually only detects the carbon element of the organic compounds and is therefore referred to as organic carbon (OC). OC concentrations usually depend on the calculation methods and measuring protocols (Cavalli et al., 2010; Chiappini et al., 2014; Zanatta et al., 2016). As a result of the different quantification, the correlation between organic matter (OM) and OC is much lower than for inorganic compounds (R²= 0.55 in PM_1.0 and R²= 0.80 in PM_2.5). Because OM also includes associated hydrogen, oxygen, and other elements, OM is significantly higher than OC, indicated by a slope of from 2.77 ± 0.92 for PM_1.0 and 2.11 ± 1.27 for PM_2.5. On average the OM $/$ OC ratios were 1.58 ± 0.54 for PM_1.0 and 1.97 ± 0.59 for PM_2.5 in this study, which are common ratios of OM $/$ OC observed in the organic aerosol. The lower ratio for PM_1.0 indicates more hydrocarbon-like aerosol at smaller particles and the higher ratio for PM_2.5 more oxidized aerosol in larger particles. Several effects could lead to inaccurate OM $/$ OC ratios and lower correlation coefficients in the data. Volatile and semi-volatile organic compounds (VOCs and SVOCs) cause positive and negative artifacts in the estimation of OC (Cheng et al., 2019; Turpin et al., 1994; Cheng et al., 2011). The positive artifact results from the adsorption of VOCs and SVOCs on quartz filters, leading to an overestimation of OC mass and thus underestimated OM $/$ OC ratios. Based on previous measurements at the same location, we estimate the upper limit of the positive artifact in this study on the order of 20 %–30 % (Dusek, unpublished data). Sometimes studies found higher artifacts up to a factor of 2, but this would lead to unrealistically high OM $/$ OC ratios in our case. Negative artifacts arise from the evaporation of SVOCs collected on the filter during sampling and potentially during storage. In order to mitigate the latter artifact, we conducted the OC–EC analysis promptly after the campaign and stored the filters in the freezer. Regarding the ACSM data, a critical factor is called the “Pieber effect”, which observed that the inorganic salts can thermally decompose and release carbonaceous material already present in the instrument, leading to the formation of CO${}_{\mathrm{2}}^{+}$ ($m/z$ 44) ions that are not related to the organic aerosol (Freney et al., 2019; Pieber et al., 2016). Data showed that the degree of interference was highly variable between instruments and over time, and CO${}_{\mathrm{2}}^{+}$ was overestimated by 0.4 % to 10.2 %. This would lead to an overestimation of OM $/$ OC ratios by up to 10 %. In the Netherlands, values towards the upper limit are more likely due to the inorganic concentrations in the Netherlands, especially of ammonium nitrate. Specifically, NH₄NO₃ resulted in a median CO²⁺ overestimate that was 3.4 % higher compared to HNO₃. The level of interference caused by other semi-refractory nitrate salts was 2–10 times higher than that caused by NH₄NO₃. In contrast, (NH₄)₂SO₄ induced interference that was 3–10 times lower than NH₄NO₃. Apart from this, a constant RIE of 1.4 was assumed for OA during the study based on the recommendation by Aerodyne, which can contribute to uncertainties in OA quantification, since this RIE can change for different instruments and different OA composition and concentration. Although there are some studies that attempted to convert the ACSM f₄₄ signal to O : C ratios and to derive OM : OC ratios, the large variability of the f₄₄ signal itself causes a large uncertainty in the O : C ratio (Crenn et al., 2015; Canagaratna et al., 2015; Aiken et al., 2008; Rollins et al., 2010; Poulain et al., 2020). Thus, this approach was not attempted in this study. In summary, OM $/$ OC ratios in this study have considerable uncertainties but are within the range of typical values found in the literature (Aiken et al., 2008; Poulain et al., 2020; Sun et al., 2011; Zhao et al., 2020).

The eBC was measured online using the MAAP with a PM₁₀ inlet, whereas the EC was collected on the filters using the PM_2.5 inlet and then analyzed offline by the sunset analyzer. Figure 1 shows the comparison of eBC and the PM_2.5 EC with a good correlation (R² of 0.83). The slope was 1.55 ± 0.44, mainly reflecting the difference in size cutoff. Moreover, it is worth noting that the MAAP instrument measures eBC at 637 nm, encompassing both BC and other light-absorbing species that share the same absorption wavelength, such as brown carbon potentially leading to overestimation of the eBC measurements (Cheng et al., 2019). Additional uncertainties are related to filter loading and multiple scattering effects (Petzold et al., 2005; Petzold and Schönlinner, 2004). The measured eBC is based on a MAC of 6.6 m² g⁻¹ for black carbon (Petzold et al., 2002) for converting the absorption to the mass concentration of eBC. In reality, this MAC value can vary widely among different environments. On the other hand, EC measurements by thermal–optical analysis (TOA) also have significant uncertainties. Previous studies show that EC can be both overestimated or underestimated by TOA depending on the thermal protocol, optical correction method, and filter loading (Yang and Yu, 2002; Schmid et al., 2001; Panteliadis et al., 2015; Cadle et al., 1980; Zenker et al., 2020), which can introduce addition uncertainties when comparing eBC and EC measurements. A recent comparison between the MAAP and OC $/$ EC analysis shows differences of 20 % for an urban site and 70 % for a regional site (Karanasiou et al., 2020). The 55 % differences found in our studies with different size cutoffs show a reasonable result.

To sum up, the comparison between the online and offline measurements of the PM_1.0 and PM_2.5 showed consistent results, especially for the SIA with slopes between 0.88–1.09 and R² values greater than 0.87. The OA vs. OC and eBC vs. EC comparisons showed results in line with previous studies. Overall, the data were fairly accurate and reliable for further study. In particular, the configuration of TOF-ACSM-CV-PM_2.5 lens showed a high stability and accuracy. With suitable inlets it can perform both NR-PM_1.0 and NR-PM_2.5 measurements for the purpose of long-term field observation.

3.1.2 Comparison of chemically derived volume concentration and PNSD-derived volume concentration

The total TOF-ACSM volume concentration was also compared and validated by the particle volume concentration derived from the PNSD. The aerosol particle size distribution with a range of around 8–800 nm in electromobility diameter was obtained by the MPSS during the RITA-2021 campaigns (in May and September) to further validate the chemical measurements. Simply put, the volume concentration from ACSM was calculated as the mass concentrations of individual species divided by the corresponding density. The density of each species using in this study was 1.75 g cm⁻³ for the inorganics (Haynes, 1942), 1.2 g cm⁻³ for the organics (Turpin and Lim, 2001), 1.52 g cm⁻³ for chloride (Haynes, 1942), and 1.77 g cm⁻³ for eBC (Park et al., 2004; Poulain et al., 2014). The MPSS volume concentration was estimated by converting the PNSD to the particle volume distribution. The total volume concentration of the MPSS is the integral of the particle volume distribution over all the size bins. Figure 2a shows the time series of the volume concentrations derived from ACSM + MAAP measurements and the MPSS-derived volume concentration. The agreement was good over the whole measurements period, indicating a stable condition of the instrument and satisfactory quality. The correlation of volume concentrations is displayed in Fig. 2b, with data points colored by the RH. The slope was nearly 1 (±0.02) with R²= 0.91, which was comparable with previous studies (Poulain et al., 2020; Pokorná et al., 2022). However, it demonstrates that the linear correlation between the two variables is significantly influenced by relative humidity. Higher relative humidity led to a lower size cutoff diameter, resulting in a lower mass concentration measured by ACSM. As also reported in the previous studies, the aerosol hygroscopic growth has a great impact on the size cut off in terms of dry particle size (Chen et al., 2018) when the ambient RH is high. It has been pointed out that the difference between ambient and dry cutoff size is approximately 10 % and 20 % for PM_1.0 and PM_2.5 in the European background and even larger in marine or coastal stations, with up to 43 % and 62 % for PM_1.0 and PM_2.5 (Poulain et al., 2020). The upper cutoff for the ACSM inlet is ∼ 2.5 µm (ambient, aerodynamic) and ∼ 0.8 µm (dry, electrodynamic equivalent) for the MPSS. Nevertheless, the dry, electrodynamic equivalent cutoff size of the ACSM inlet will be larger than 0.8 µm. Therefore, the ACSM volume concentrations were expected to be higher, and it is surprising that the agreement is so close. However, the ACSM only measures non-refractory material, whereas the MPSS-derived volume concentration also includes non-refractory material. This indicates that there is considerable contribution from non-refractory material other than BC. The filter analysis also supports this conclusion, as seen in Fig. S5, which shows approximately 21 % of the PM_2.5 mass was not detected by the ion analysis. Thus, the slope of 1.00 is probably a coincidence, where missing volume from the MPSS cutoff and missing mass from the ACSM roughly cancel out. On the whole, the high R² values give confidence in the stability and accurateness of the ACSM instrument in the long-term observations. A comparison between ACSM and MPSS volume concentrations is highly recommended as a regular quality control strategy.

Figure 2(a) The time series of the ACSM and MAAP volume concentrations (red line) compared with the MPSS volume concentration (blue area). (b) The linear regression fitting correlations of the ACSM and MAAP volume concentration with the MPSS-derived volume concentration. Scatters colored by the relative humidity (%).

Download

3.2.1 Comparison of NR-PM_1.0 and NR-PM_2.5 species measured by TOF-ACSM

As mentioned, two identically configured TOF-ACSMs with PM_2.5 aerodynamic lens were collocated and set up to measure the NR-PM_1.0 and NR-PM_2.5 during the RITA-2021 spring campaign. At the start of the campaign, both instruments were intercompared by measuring the NR-PM_2.5. The results shown in Fig. S1 demonstrate good comparability, with the R² ranging from 0.91 to 1.0 and slopes ranging from 0.94 to 0.99 for the SIA compounds SO₄, NO₃, and NH₄. The slopes were not significantly different from 1 at the 95 % confidence level. For chloride, the correlation was not as good as for other species because ammonium chloride had a very low concentration during the whole measurement period. Therefore, it will not be discussed in the following. The correlation of PM_1.0 and PM_2.5 OA concentrations was also reasonable with a slope of 0.93 ± 0.13 (R²= 0.80). Overall, the two collocated TOF-ACSMs compared well and can be used to compare PM_1.0 and PM_2.5 chemical composition.

Figure 3 shows the total mass concentration time series of the NR-PM_1.0 and NR-PM_2.5, as well as the concentration of individual chemical species and the corresponding scatter plots with regression lines. The mass concentration of NR-PM_2.5 was on average 5.27 ± 3.98 µg m⁻³, with a range from 4.84 to 22.25 µg m⁻³. This concentration was below the WHO PM_2.5 annual limit values (10 µg m⁻³) (World Health Organization, 2021) and also lower than previously reported concentrations in this region of 14.4 ± 2.1 µg m⁻³ (Schlag et al., 2016; Mensah et al., 2012; Mamali et al., 2018). The PM_1.0 and PM_2.5 mass concentrations of each species were highly correlated over the whole measurements period with R²≥ 0.98. In general, the PM_1.0 SIA accounted for approximately 75 %–85 % of the PM_2.5 SIA on average, with individual contributions ranging from 82±1 % for ammonium and 79±1 % for nitrate to 76±1 % for sulfate. For organics, the PM_1.0 accounted for a higher fraction of PM_2.5, with around 85±1 %. Similar results were also found in the filter samples as displayed in Fig. S2. In addition, EC-PM_1.0 accounted for 74±14 % of the EC-PM_2.5. In general, the PM_1.0 mass concentration explained 80±1 % of the PM_2.5 on average, and this ratio ranged from 45.21 %–94.78 % throughout the campaigns. However, there was still a substantial proportion (∼ 21 %) of unexplained mass in the PM_2.5 as shown in Fig. S5.

Figure 3Time series of the NR-PM_1.0 and NR-PM_2.5 chemical species and the total mass concentration measured by TOF-ACSM and corresponding linear regression fitting correlations.

Download

In addition, the chemical mass fractions of PM_1.0 and PM_2.5 displayed in Figs. S3–S4 revealed that there were some slight differences in the chemical composition of the PM_1.0 and PM_2.5. Figure S3 showed the average hourly mass fraction measured by the ACSM for the NR-PM_1.0 and NR-PM_2.5. The OA accounted for similar proportions, namely 34.4 % of NR-PM_1.0 and 33.0 % of NR-PM_2.5. NO₃ contributed 27.8 % to NR-PM_1.0, with a slight increase to 31.5 % in NR-PM_2.5. Figures S4 and S5 show the daily and the average mass fractions for PM_1.0 and PM_2.5 species from the filter samples, with a higher NO₃ fraction in PM_2.5 and a lower OC fraction in PM_1.0 for the whole period. Specifically, the NO₃ fraction increased from 38.3 % in PM_1.0 filter samples to 45.5 % in PM_2.5 filter samples, whereas the OC fraction decreased from 19.9 % to 15.1 %. The difference between the ACSM OA mass fractions (similar in PM_1.0 and PM_2.5) and the OC mass fraction on the filters (higher in PM_1.0 than in PM_2.5) is the result of higher OM $/$ OC ratios for larger particles. As discussed in Sect. 3.1.1, this is likely due to the fact that pure hydrocarbons that are often contained in primary emissions are more abundant in the smaller particle size range. This change in chemical composition with particle size suggests that different types of particles may dominate in different size ranges, potentially indicating a non-internal aerosol mixing state during the measurements. The differences of OC mass fraction in PM_1.0 and in PM_2.5 also further explain the stronger correlation of ACSM OA and Filter OC in PM_1.0 compared to in PM_2.5 shown in Fig. 1.

Figure 4Time series of TOF-ACSM-measured SIA-PM_2.5 during the whole period. MARGA-measured PM_1.0 from 5 to 30 September 2021 and PM_2.5 from 3 to 15 October 2021. The corresponding linear regression fitting correlations between MARGA PM_1.0 and TOF-ACSM PM_2.5.

Download

3.2.2 Comparison of the SIA-PM_1.0 by MARGA and SIA-PM_2.5 by TOF-ACSM

The comparison of the MARGA and ACSM measurements was carried out for further evaluation and validation. Figure 4 displays the time series of the MARGA measurements and TOF-ACSM measurements. Figure 5 compares MARGA and TOF-ACSM data during time periods when both instruments measured PM_2.5. Strong correlations with R² ranging from 0.93 to 0.97 and small intercepts demonstrated a good reliability of the two methods. However, the linear regression slopes display some discrepancies for individual species. The NO₃ measured by the ACSM and MARGA showed an excellent agreement, with a difference below 3 % (slope $=\mathrm{0.97}\pm \mathrm{0.03}$, R²= 0.97). For the NH₄ and SO₄, the MARGA mass concentrations were lower than the ACSM mass concentrations. The slope for NH₄ was 0.83 ± 0.04, and for SO₄ it was 0.78 ± 0.02. The analysis also revealed a dependence on the RH for the correlation between the two measurements. As illustrated in Fig. 5, the ACSM tended to measure more mass than the MARGA under lower RH conditions. The hygroscopic growth of the aerosol particles at higher RH resulted in lower dry cutoff sizes, and the different inlets of the MARGA and ACSM might lead to differences in the detected mass. Combined with Fig. S6, it shows a slight bias between ACSM and MARGA at higher concentrations, particularly when pollution originates from the south or southeast. Since the ACSM mass also includes contributions from organic nitrates, organic sulfates, and organic ammonium salts to the observed NH₄, NO₃, and SO₄ concentrations, this could also lead to higher concentrations observed by the ACSM. However, given that the validation of the TOF-ACSM against filter samples showed excellent agreement for these ions as discussed in the Sect. 3.3.1 and listed in Tables S1–S3, the ACSM results are more likely to be closer to the true values compared with MARGA. Most previous comparisons of ACSM/AMS and MARGA showed that the MARGA gave higher concentrations when the ACSM/AMS used the PM_1.0 lens (Zhao et al., 2020; Stieger et al., 2018; Heikkinen et al., 2020). To the best of our knowledge, this is the first comparison between the PM_2.5 lens on a TOF-ACSM-CV and a MARGA. We observed that a higher concentration can be achieved using the CV and PM_2.5 lens of the TOF-ACSM, which further verifies its capability in measuring non-refractory PM_2.5 and PM_1.0 concentrations quantitatively. Because of the very high correlation between MARGA and ACSM concentration, the slight bias between the instruments can be corrected using the regression coefficients in Fig. 5. Figure 4 also displayed the linear regression correlations between the MARGA-measured PM_1.0 and TOF-ACSM- measured PM_2.5 inorganic chemical species after this correction. The correlations between ACSM PM_2.5 and MARGA PM_1.0 all showed R² values greater than 0.85, and the slopes are 0.76 ± 0.03 for NH₄, 0.74 ± 0.03 for SO₄, and 0.70 ± 0.02 for NO₃, very comparable to the slopes achieved in the spring campaign, using two different ACSMs. In summary, the local concentrations of both PM_1.0 or PM_2.5 were relatively low throughout the observation period. The PM_1.0 and PM_2.5 studied by using the several different instruments have demonstrated that the PM_1.0 mass concentrations accounted for 70 %–80 % of the PM_2.5 concentrations for various non-refractory species.

Figure 5The linear regression fitting correlations between MARGA PM_2.5 and TOF-ACSM PM_2.5, with points colored by the relative humidity (%). The blue shaded area represents the 95 % confidence interval.

Download