What chemical species are responsible for new particle formation and growth in the Netherlands? A hybrid positive matrix factorization (PMF) analysis using aerosol composition (ACSM) and size (SMPS)

. Aerosol formation acts as a sink for gas-phase atmospheric species that controls their atmospheric lifetime and environmental effects. To investigate aerosol formation and evolution in the Netherlands, a hybrid positive matrix factorization 15 (PMF) analysis has been conducted using observations from May, June, and September 2021 collected in a rural site of Cabauw in Central Netherlands. The hybrid input matrix consists of the full organic mass spectrum acquired from a time-of-flight aerosol chemical speciation monitor (ToF-ACSM), ACSM species concentrations, and binned particle size distribution concentrations from a scanning mobility particle sizer (SMPS). These hybrid PMF analyses discerned six factors that describe aerosol composition variations: four size-driven factors that are related to new particle formation and growth (F6, F5, F4, and 20 F3), and two bulk factors driven by composition, not size (F2, F1). The smallest-diameter size factor (F6) contains ammonium sulfate and organics, and typically occurs during the daytime. Newly formed particles, represented by F6, are correlated with wind from the southwesterly-westerly, northerly, and easterly sectors that transport sulfur oxides (SO x ), ammonia (NH 3 ), and organic precursors to Cabauw. As the particles grow from F6 to F3, nitrate plays an increasing role, and the particle loading diurnal cycle shifts from daytime to a nighttime maximum. The inorganic ion balance and organics composition in the bulk 25 atmosphere affects the chemical composition variation across factors and seasons. Changing ammonium-sulfate-nitrate equilibrium shifts inorganic species among factors, and greater organics availability makes secondary organic aerosol (SOA) more influential in summertime aerosol growth, principally due to volatility differences produced by seasonal variation in photooxidation and temperature.

Aerosols impact the Earth by absorbing and scattering solar and terrestrial radiation (Andreae and Crutzen, 1997;Grantz et al., 2003;Wong et al., 2017;Marrero-Ortiz et al., 2019), and indirectly by producing or modifying clouds (Lohmann and Feichter, 2005;Mahowald et al., 2011;Fan et al., 2018).NPF plays a prominent role in cloud formation by contributing to over 50% of cloud condensation nuclei formation, which affects the lifetime and radiative properties of clouds (Bianchi et al., 45 2016;Gordon et al., 2016;Haywood, 2016;Dall'Osto et al., 2018;Lee et al., 2019).These phenomena affect the ecosystem physically by modifying radiation diffusion, temperature, and precipitation (Grantz et al., 2003;Haywood, 2016;Lee et al., 2019).Aerosols also influence the ecosystem chemically through influencing the spatial patterns of nitrogen deposition (van der Swaluw et al., 2011;Wamelink et al., 2013) and oxidative processes (Xing et al., 2017), leading to ecological harm such as soil pollution, water acidification, eutrophication, and loss of biodiversity (Erisman et al., 2011;Wamelink et al., 2013).In 50 terms of public health, aerosols exhibit adverse effects on human health due to their size and chemical composition.NPF events are typically followed by air quality degradation, which is consistently associated with elevated pulmonary and cardiovascular morbidity and mortality worldwide (Ayala et al., 2012;Pope et al., 2020).Sulfuric acid (H2SO4) is typically understood to be the most prevalent nucleation-inducing agent in NPF events, together with 55 other airborne chemical species, including nitrate, bases (e.g., amines), and organic acids (Zhang et al., 2012;Kulmala et al., 2013;Zhang et al., 2015;Wagner et al., 2017;Lehtipalo et al., 2018;Kürten, 2019;Lee et al., 2019;Brean et al., 2021;Olin et al., 2022).Numerous studies also report low-volatility organic species, such as terpene oxidation products and organic nitrates, participating in the formation of new particles (Berkemeier et al., 2016;Bianchi et al., 2016;Tröstl et al., 2016;Barsanti et al., 2017;Dall'Osto et al., 2018;Kerminen et al., 2018;Lee et al., 2019;Heinritzi et al., 2020).60 In this work, we show that co-located measurements of aerosols' atmospheric composition and particle size distribution can be used to characterize the chemical composition of new particle and aerosol components that facilitate their growth.A time-offlight aerosol chemical speciation monitor (ToF-ACSM, Aerodyne Inc.) allows the continuous and real-time quantification of non-refractory chemical species in ambient air (Ng et al., 2011;Fröhlich et al., 2013).For particle size distributions, the 65 scanning mobility particle sizer (SMPS) provides real-time measurement of submicron particle number concentrations of different sizes (Amaral et al., 2015;Wiedensohler et al., 2018).
Aerosol mass spectrometry measurements have been used extensively with positive matrix factorization (PMF) as a strategy for aerosol source apportionment, especially regarding the organic components (Lanz et al., 2007;Jimenez et al., 2009;Ulbrich 70 et al., 2009;Ng et al., 2010;Zhang et al., 2011).This paper combines the full organic mass spectrum and chemical species concentrations from ToF-ACSM with particle size distribution from SMPS into a hybrid PMF input matrix, to study the association between chemical composition and particle size distribution.A similar approach for hybrid ACSM-SMPS PMF analysis was used for a European aerosol dataset comparison (Dall'Osto et al., 2018).Previous studies on aerosol source apportionment in the Netherlands have focused on organic aerosol composition (Mooibroek et al., 2011;Mensah et al., 2012;75 Schlag et al., 2016).Here, we analyze ACSM-SMPS datasets from Cabauw, the Netherlands, collected as part of the Ruisdael Observatory Land-Atmosphere Interactions Intensive Trace-gas and Aerosol (RITA) campaign in May to September 2021 (https://ruisdael-observatory.nl), using PMF to characterize the chemical species responsible for new particle formation and growth across several seasons.Several studies have shown NPF events dependent on air mass origin transporting different pollutants (Hamed et al., 2007;Modini et al., 2009;Castro et al., 2010;Asmi et al., 2011;Németh and Salma, 2014;Nieminen 80 et al., 2014;Qi et al., 2015;Mordas et al., 2016;Kolesar et al., 2017;Peng et al., 2017;Kerminen et al., 2018;Pushpawela et al., 2019), and therefore we also explore relationships between wind direction, wind speed, and factor timeseries to interpret source apportionment. of Noord-Brabant.In summer (June), the prevailing air masses were coming from west-northwest to north-northeast (292.5° to 22.5°), bringing air from the North Sea and some major cities along the coast and/or in the Randstad, such as Rotterdam, The Hague, Amsterdam, and Utrecht.More diverse wind plumes were observed in September, ranging from easterly (22.5° to 112.5°), coming from the forested nature and agricultural areas in the province of Gelderland, and southerly (180° to 202.5°), coming from the province of Noord-Brabant.The meteorological variables for each period are summarized in Table 1 and the 110 wind variables are visualized as wind roses in Fig. 2. In June, the prevailing winds were from WNW up to the NNE sector.In September, two major wind directions were from the E and S sectors.

Measurements
The ToF-ACSM was the main instrument employed for this analysis.The instrument has been detailed in other work (Fröhlich 120 et al., 2013).The ACSM was installed with an inlet at 5-meter height, a cyclone with size cut of 2.5 µm (PM2.5), and an aerodynamic lens in the inlet system allowing the analysis of non-refractory organics, ammonium, nitrate, sulfate, chloride, and potassium in the aerosol phase.The instrument is equipped with a capture vaporizer (CV) instead of the standard vaporizer (SV), which is designed to increase the collection efficiency (Middlebrook et al., 2012;Jayne and Worsnop, 2016).By having a narrow entrance, the CV increases the particle collision events and thus increases the contact with the hot vaporizer surface, 125 minimizing particles that bounce without evaporation (Hu et al., 2017).Consequently, however, the fragmentation patterns are shifted towards smaller ion masses due to the additional thermal decomposition (Hu et al., 2017;Zheng et al., 2020).The ToF-ACSM provides unit mass resolution (UMR) mass spectra which are analyzed using Tofware v3.2 in Igor Pro 8.The fractions of measured UMR signals were assigned to individual aerosol species using the fragmentation table.On-site calibrations are performed to determine the ionization efficiencies of the chemical species.The calibration with ammonium nitrate and 130 ammonium sulfate gives ionization efficiency (IE) value of 250.0 ions pg -1 for nitrate (NO3), and relative ionization efficiency (RIE) values of 1.40, 1.67, 1.30, 3.35, and 1.00 for organics, sulfate (SO4), chloride (Cl), ammonium (NH4), and potassium (K), respectively.
In addition to the aerosol measurements by the ToF-ACSM, ambient sulfur dioxide (SO2) concentrations were obtained from 135 the open-source data of Landelijk Meetnet Luchtkwaliteit (LML, https://www.luchtmeetnet.nl),measured at the same location.Ammonia (NH3) concentrations were obtained from measurements in Zegveld-Oude Meije station, 20 km to the north of Cabauw station (see Fig. 1), also acquired from LML.The particle size distribution concentration measurements were conducted using a TROPOS-SMPS with 8 nm to 800 nm size range and 5-minute time resolution from a 5-meter height inlet.
The co-located weather data were retrieved from the Royal Netherlands Meteorological Institute (KNMI, 140 https://www.knmi.nl).
The errors for each size bin are taken to be the population standard deviation of the raw data.The variable values of species mass concentrations and particle number concentrations were weighted prior to analysis to ensure that individual peaks have similar magnitude as the organic mass spectrum.This weighting was done to give all variables the same importance in the PMF analysis.155 We performed the analysis using the PMF Evaluation Tool (PET) v3.08 (Ulbrich et al., 2009) in Igor Pro 8.The details of applying positive matrix factorization (PMF) to aerosol mass spectrometry datasets have been discussed elsewhere in detail (Paatero and Tapper, 1994;Paatero, 1999;Ulbrich et al., 2009).The first step of the factor analysis was identifying the optimum number of factors by running unconstrained experiments using 2 to 10 factors.The optimum number of factors is selected by 160 considering the lowest residuals of the PMF solutions, and whether all the factors are environmentally reasonable and unique based on their chemical composition.For all three-month datasets, this analysis converges to an optimum of 6 factors.The ordering of the factors is not identical across months and for ease of presentation, we reorder them so that factors of similar identity have the same factor number across months.The percent of signal explained by each factor is presented alongside plots of the factor profiles, to enable the reader to deduce the original factor ordering.The resulting hybrid PMF solution matrix 165 is split back up into organic mass spectrum, species mass concentrations, and particle number concentration bin matrices, and the value of each variable is readjusted (undoing the previously described re-weighting) so that the ACSM species mass concentrations can still be interpreted quantitatively.

Wind analysis
To analyze the factors using wind variables, we investigate the prevailing wind for several pollution episodes observed in the 170 dataset.Bivariate polar plots are generated for selected factor reconstructed mass concentration derived from PMF analyses and mass concentration for each ACSM species in each period using the "Openair" package in the "R" environment (Carslaw and Ropkins, 2012).The wind parameters are obtained from co-located measurement of 10-meter wind direction data acquired from KNMI.

Factor particle size distributions and composition
From the unconstrained experiments, the best PMF solution was found with six factors for the combined ACSM-SMPS matrix in each of the datasets for May 2021 (Fig. 3), June 2021 (Fig. S1), and September 2021 (Fig. S2), with similarities observed across months.As each period encompasses around one month of time series data, the factors discerned by the hybrid PMF 180 analyses show typical average aerosol composition during each period, rather than individual pollution episode profiles that may vary over time.
Four factors have particle size profiles distributed in specific diameter subranges, which we interpret as related to new particle formation and growth.We therefore call these factors "size-driven".The size-driven factors resolved from the analysis possess https://doi.org/10.5194/egusphere-2023-554Preprint.Discussion started: 5 April 2023 c Author(s) 2023.CC BY 4.0 License.some similarities in composition across months, from ammonium sulfate-rich aerosol (F6) to varying mixture of organic aerosol (OA) and inorganic aerosol (IA) (F5, F4, and F3) (see Fig. 4).We interpret the smallest-diameter size-driven factor as a nucleation-mode factor (F6), which then grows with different composition, on average progressing from F6 to F3.The two bulk composition-driven factors are distributed across particle sizes, consisting of: (F2), an OA and IA mixed factor with a variable composition, and (F1), an organic aerosol factor with occasional traces of ammonium sulfate aerosol.190

Factor organic profiles
The organic mass spectrum of each factor profile can be used to obtain information regarding the oxidation level of the factor.
Oxidized organic compounds are characterized by a relatively high m/z 44 (f44) signal, originating primarily from CO2 + fragments of carboxylic groups in organic compounds produced by thermal decomposition inside the ACSM vaporizer (Alfarra et al., 2004).This f44 fragment is often related to a high degree of oxidation and photochemical ageing (Alfarra et al., 2004;195 Ng et al., 2010).The m/z 43 (f43) fragment is characteristic for both oxidized organic compounds (CH3CO + ) and saturated hydrocarbon compounds (C3H7 + ).Thus, factors with higher f44 and lower f43 values are understood to be more oxidized and are often classified as an oxygenated organic aerosol (OOA).Meanwhile, lower f44 and higher f43 values implies that the factor is less oxidized and is often labelled hydrocarbon-like organic aerosol (HOA).In this paper, we use these values only to compare the oxidation level between aerosol factors.It is also important to note that the ACSM used in this paper has a CV 200 instead of a SV inlet (see Sect. 2.2), which is known to produce higher f44 values due to enhanced thermal decomposition (Hu et al., 2017), and therefore cannot be directly compared to PMF organic profiles in other works.

Bulk atmospheric chemical composition across periods
We hypothesize that the bulk atmospheric chemical composition influences how the chemical species are distributed across the PMF factors.The three periods have different average chemical composition as detected by the ToF-ACSM measurements.
The mean concentrations of atmospheric species and the species percentages in the bulk atmosphere are summarized in Table S1.For this assessment, we are using three mean concentration ratios summarized in Table 2 to characterize average 220 composition.Ion balance ratio (n NH 4 /(n NO 3 +2×n SO 4 +n Cl )) is the ratio between the measured ammonium (n NH 4 ) and the total ammonium required to neutralize the major anions (n NO 3 +2×n SO 4 +n Cl ).It illustrates the excess of atmospheric ammonium (cation), or nitrate (anion), and other possibilities based on aerosol chemistry (see Sect.S2 for details).This ion balance ratio was found to be unity in May (suggesting charge-balanced condition), less than unity in June (suggesting nitrate excess), and more than unity in September (suggesting ammonium excess).We also examine the mean sulfate-to-nitrate molar 225 concentration ratio ( n SO 4 /n NO 3 ), determining a sulfate-rich or sulfate-poor condition relatively.The mean organic-toammonium mass ratio (m Org /m NH 4 ) is used to show whether a period is organic-rich relative to other periods.
We have three different composition regimes measured during the three periods based on these ratios.In springtime (May), we can see a sulfate-rich and nitrate-excess regime.In summertime (June), we have an organic-rich, sulfate-rich with nitrate-230 excess regime.Lastly, a sulfate-poor and ammonium-excess regime is observed in autumn (September).Below, we will discuss how these regimes influence the chemical composition of factors that were obtained from PMF analyses.
Table 2. Mean bulk atmospheric chemical composition in the three periods, summarized as the values of ion balance ratio ( n NH 4 /(n NO 3 +2×n SO 4 +n Cl ) ) from linear regression, mean sulfate-to-nitrate ratio ( n SO 4 /n NO 3 ), and mean organic-to-ammonium ratio 235 (m Org /m NH 4 ).The concentrations were detected by the ToF-ACSM in Cabauw.A more detailed information can be seen in Table S1.

Mean ratio
May

Size-driven factors (F7, F6, F5, and F4)
Using the hybrid ACSM-SMPS datasets, four size-driven factors emerge from the PMF analyses as F6, F5, F4, and F3 (see Fig. 3, Fig. S1, and Fig. S2).These factors are considered "size-driven" due to the approximately normally distributed particle 240 concentrations in a specific sub-range of diameter.The four factors display different particle size clusters increasing in diameter from F6 to F3.The size-driven factors have diverse composition characterized by a variety of OA and IA mixtures, as can be seen in Fig. 4, discussed in more detail below.New particle formation (NPF) events are characterized by the rapidly increasing particle number concentration below 20 nm followed by particle growth, creating nearly vertical aligned peaks in particle number concentration plotted against time (Heintzenberg et al., 2007;Kerminen et al., 2018).In Fig. 5 and Fig. S3, we observe that the episodes during which the smallest size-driven factor (F6) fraction increases occur when the total aerosol mass concentration is relatively low.If we zoom into the timeseries, the NPF growth shapes appear during episodes that are dominated by the smallest size factor, F6 (see Fig. S4).255 The trace inorganic ions (chloride and potassium) in the factors are found to be correlated with the existence of sulfate and organic species, respectively, hinting at possible correlations with sea spray and biomass burning aerosol, respectively (see details in Sect.S4).

F6: nucleation-mode factor
F6 correspond to particles in the nucleation mode growing into Aitken mode size range (8 < Dp < 65 nm).The size range differs slightly across months but is always the smallest particle size range among factors, which we therefore designate as nucleation mode.In all periods, F6 has ammonium sulfate as a major component.The factor further consists of ammonium 270 (7.1% to 32.4%), sulfate (20.3% to 61.6%), organic compounds (0.0% to 70.4%), and traces of chloride (0.1% to 6.0%) and potassium (0.0% to 2.1%).
The mass percentage share between ammonium, sulfate, and organics of F6 depend on the composition regime in each period (see Fig. 4).The sulfate-rich regime with low bulk organics in springtime (May) leads to F6 being largely ammonium and 275 sulfate (19.1% and 54.2%, respectively), followed by OA (22.4%) and chloride (4.8%).The organic-rich and sulfate-rich regime in summer (June) results in F6 being dominated by OA (70.4%) rather than ammonium and sulfate (7.1% and 20.3%, respectively), with traces of chloride (0.1%) and potassium (2.1%).In autumn (September), the ammonium-excess and sulfatepoor regime with low bulk organics results in the F6 composition being exclusively ammonium and sulfate (32.4% and 61.1%, respectively), with traces of chloride (6.0%) and potassium (0.5%).We interpret these results to mean that sulfate is a key 280 component of condensation nuclei during NPF events, regardless of the bulk atmospheric composition.When the bulk organic concentration is high, it participates in particle nucleation.With excess ambient ammonium, however, particle condensation can occur without organic compounds.
Oxidation levels further explain the OA contribution difference in nucleation-mode particles.We can observe this variation 285 by looking at the organic fragment spectrum of F6.High f44 value is observed in F6 where OA participates in new particle condensation during spring (May, f44 = 0.32) and summer (June, f44 = 0.33).In these months, higher mean radiation promotes photooxidation of organic vapors and semi-volatile oxidized organic compounds condense rapidly onto newly formed particles, resulting in higher OA contribution to nucleation-mode particles.In September, when mean radiation is lower, OA does not appear in F6.The organic mass spectra are consistent with this lower degree of oxidation, with lower f44 and high f43 290 value in autumn (September, f44 = 5.8×10 -6 and f43 = 0.16).We infer that only lower volatility, oxidized organic compounds can condense on freshly nucleated particles.The organic-rich regime combined with higher mean temperature favors the abundant condensation of available semi-volatile OA onto newly formed particles.Consistent with these results, several aerosol chamber experiments have reported that highly oxygenated-295 organic molecules (HOM) from biogenic and anthropogenic organic precursors play a dominant role in new particle formation and growth (Schobesberger et al., 2013;Ehn et al., 2014;Riccobono et al., 2014;Tröstl et al., 2016;Pospisilova et al., 2020;Zhao et al., 2021).
The size range differs across period but has particle size progressively growing from F5 to F3, which we thus designate the "growth-mode" factors.The growth-mode factors contain different composition across seasons, depending on the available semivolatile species for condensation.

305
In springtime (May), less-volatile and less-oxidized organic compounds (f44 = 8.2×10 -6 and f43 = 0.13) condense onto newly formed particles, increasing the organic contribution in F5.The increase of OA in F5 is also observed for autumn (September), although in the case more highly oxidized organic compounds (f44 = 0.21 and f43 = 0.07) condense onto new particles.We suggest that although oxidized organic compounds are not involved in early particle condensation when the mean radiation is lower, they are incorporated in later growth.For spring and autumn, F5 consists of largely OA composition (84.0% to 87.4%), 310 ammonium (4.3%only in May), chloride (8.3% to 9.0%), and potassium (2.7% to 4.3%).
In summertime (June), there is a substantial amount of nitrate (52.4%) in F5 alongside less oxygenated OA (28.4%, f44 = 2.8×10 -8 and f43 = 0.17), ammonium (16.8%), chloride (0.9%), and potassium (1.5%).The ion balance ratio in June was less than unity, which can be due to excess nitrate aerosol (see Sect.S2).We propose that it indicates the promotion of semi-volatile 315 organic nitrate formation in the summertime, condensing onto the newly formed particles as the temperature lowers.Organic nitrate comes from reaction between NOx with less oxygenated organic compounds and condensing into aerosol phase more easily in this period.Further, we infer the possibility of biogenic source of the organic compounds involved in this organic nitrate source, due to the fact that we observe this composition only in the hottest and sunniest month of June and less

oxygenated. 320
As particles grow into the larger size ranges F4 and F3, they incorporate more OA and ammonium salts.F4 resembles the composition of either the two smallest size-driven factors (F6 or F5), which may be due to continuation of smaller particle growth with the same composition.F3, the largest particle size range among size-driven factors, contains nitrate as the main component (43.0% to 48.0%) followed by OA (26.2% to 36.0%), and ammonium (14.4% to 18.1%).The dominant 325 contribution of nitrate is observed across periods regardless of the nitrate composition regime.Overall, the composition of these size-driven factors suggests that, while sulfate is key to nucleation, nitrate plays a more important role in particle growth, which is related to the ammonium-sulfate-nitrate aerosol chemistry (see Sect.S2).

Relationships with mean radiation and temperatures
The reconstructed PMF masses show the influence of mean downward shortwave radiation and temperature on NPF events.330 The average PMF mass fraction of nucleation-mode particles (F6) is higher in summertime (June) compared to other periods (see Fig. S1) due to higher mean radiation and temperatures (see Table 1).In summer (June), F6 accounts in average 22.3% https://doi.org/10.5194/egusphere-2023-554Preprint.Discussion started: 5 April 2023 c Author(s) 2023.CC BY 4.0 License. of total reconstructed PMF mass while in spring (May) and summer (September), they only represent 6.2% and 4.1%, respectively.The more frequent appearance of NPF growth events during summer can be seen in Fig. S3a.The occurrence of NPF events is generally favored in high radiation (Modini et al., 2009;Peltola et al., 2022) and warmer temperatures (Jokinen 335 et al., 2022;Peltola et al., 2022).Solar radiation provides the UV radiation that promotes photochemical reactions and turbulent motions needed to form new particles (Wehner et al., 2015;Dada et al., 2017;Kerminen et al., 2018;Sellegri et al., 2019).
The diurnal cycles of the size-driven factors shown in Fig. 6 (May) and Fig. S5 (June and September) also shows the influence of shortwave radiation and temperature on particle formation.Going from F6 to F3, we observe that the factors' diurnal pattern 340 tends to shift from a daytime peak (F6 peaks in the middle of the day) to higher nighttime concentration as the particle size grows.Regional-scale NPF normally takes place during the daytime between sunrise and sunset, where it can carry on for several hours (Hamed et al., 2007;Hussein et al., 2009;Németh and Salma, 2014;Qi et al., 2015;Dai et al., 2017;Kerminen et al., 2018;Kalkavouras et al., 2020), although there are also instances of NPF happening during the night in various locations (Hirsikko et al., 2012;Vehkamäki and Riipinen, 2012;Rose et al., 2018;Kammer et al., 2018) or quiet NPF during non-event 345 days (Kulmala et al., 2012;Dada et al., 2018;Kulmala et al., 2022).Daytime peaks in F6 can be seen in spring (see Fig. 6) and summer (see Fig. S5) when the mean radiation is higher.This peak is less visible during autumn when mean radiation is lower (see Fig. S5).The sulfate-rich new particle factor F6 also appears around the same time as the precursor gas SO2, with the most obvious match between SO2 and the F6 diurnal cycle observed in spring (May, see Fig. 6).

350
Unlike the nucleation-mode F6, the growth-mode F5 and F4 have different diurnal patterns across periods, corresponding to their diverse compositions.F5 and F4 also exhibit consistently less diurnal variation, except for F5 during summer (see Fig. S5), when it tracks with the highest mean radiation across periods.In summertime (June), when there was more daylight and radiation available to promote NPF and growth, F5 has a pronounced peak in the afternoon, later than the F6 peak.

355
F3 shows more elevated nighttime concentrations across seasons (see Fig. 6 and Fig. S5).Lower nocturnal temperatures foster the condensation of semivolatile chemical species into the aerosol phase, increasing the nighttime particle concentration.The increase in the particle concentration during the night can also be facilitated by a shallower nocturnal boundary layer.

Relationship of new particle formation with wind variables
To study the relationship between wind variables and new particle formation in Cabauw, we compare timeseries of wind 365 variables and the reconstructed PMF fraction of the nucleation-mode (F6) and first growth-mode (F5) in Fig. 7 (May) and Fig. S6 (June and September).We observe that new particle formation episodes are correlated mainly with air masses transported from southwesterly-westerly sector or northerly-easterly sector.These wind sectors are the suppliers of organics, sulfate, ammonium, and their precursor gas that determine the main composition of F6 (see Fig. S7 and Fig. S8).Westerlies represent https://doi.org/10.5194/egusphere-2023-554Preprint.Discussion started: 5 April 2023 c Author(s) 2023.CC BY 4.0 License.a source of sulfate, which mainly comes from sulfur oxides (SOx) emission along the waterway of Rotterdam's harbor, the 370 busiest port in Europe.The sulfate in air transported to Cabauw from the northern to eastern sector may arise from SOx precursor from other urban, shipping, industry centers (e.g., Amsterdam city and port, Utrecht city), and power plants to the site (see Fig. 1) (Henschel et al., 2013;Fioletov et al., 2016;Ledoux et al., 2018).The supply of ammonium for new particle condensation through NH3 emission comes from the agricultural practices that take place around the Cabauw site, with tendency of higher NH3 and ammonium from the southern sector.The easterlies extending to north are sources of VOCs 375 coming from the forested nature areas in the provinces Utrecht and Gelderland, which are subsequently transformed into SOA.
The different prevailing wind affects the F6 composition and frequency.In spring (May) and autumn (September), new particle formation mainly correlates with winds from southerly and westerly directions (see Fig. 2), and thus has less organic composition.In summer (June), winds coming from the north or east contribute to NPF events (see Fig. 2), supplying more 380 organics to the site.As previously discussed in Sect.3.3.3,the abundant organics combined with higher radiation and temperatures allow semi-volatile oxidized organics to directly condense onto newly formed particles, increasing NPF events and F6 mass fraction during summer (see Fig. S6).We observe that new particle formation episodes are correlated with air masses from the southwesterly-westerly sector.Similar figures for June and September 2021 can be found in Fig. S6, showing more contributions from northerly-easterly wind directions.

Bulk composition factors (F2 and F1) 390
The two bulk composition factors yielded by the PMF analyses are F2 and F1 (see Fig. 3, Fig. S1, and Fig. S2).We call these factors bulk because they are composition-driven and found across the size distribution, rather than in a specific size range.
They collectively account for a large fraction of total aerosol mass loading (31.1% to 64.4%).
F2 is characterized by the presence of a mixture of OA and IA.In Fig. S9, we can see that the diurnal cycle of F2 is similar to 395 the diurnal pattern of total aerosol mass loading in spring (May).It is less similar in summer (June) and autumn (September) due to scarcity of OA mass, which is mainly found in the size-driven factors.This pattern suggests that F2 is the result of condensation of available semi-volatile chemical constituents over the course of the day.
F1 is composed mainly of OA with a trace amount of IA, containing the majority of OA in every month.The abundance of 400 m/z 44 fragments indicates that factor F1 represents aged organic aerosol, resembling OA profiles observed at the same site in previous studies (Mensah et al., 2012;Paglione et al., 2014;Schlag et al., 2016).The factor is related to airmasses arriving from the easterly to southerly sector (see Fig. S10 in Sect.S3 for details).Considering the organic profile, the high mass loading percentage, and the source regions across periods, we attribute F1 to background regional and continental OA.

Conclusions 405
In this work, we have shown that hybrid ACSM-SMPS PMF analysis can be used to determine the chemical constituents associated with new particle formation and growth.The analyses of three selected periods enable us to use the seasonality of the factor profiles, representing conditions of spring (sunny and warm) for May, summer (very sunny and hot) for June, and autumn (less sunny and warm) for September, as well as different prevailing winds, to attribute factor sources.Different chemical composition regimes were observed across these seasons, which manifested as differences in the species controlling 410 aerosol growth.
New particle formation episodes appeared when the total aerosol concentration was low, with key contributions from ammonium sulfate, regardless of bulk aerosol composition regime.This new particle formation and growth exhibits a diurnal pattern dominated by daytime formation that shifts to nighttime growth as the particle size increases.While sulfate promote 415 new particle formation, nitrate is more influential in condensational growth.Organics participate in either new particle formation or growth.
The abundance of organics, sulfate, and ammonium aerosol, as well as the mean radiation and temperature, govern the chemical species distribution across the factors.As result, different modes of new particle formation and growth exist.The sulfaterich regime in spring results in the nucleation-mode containing a mixture ammonium sulfate and organics.The organic-rich regime, higher mean radiation, and higher mean temperature in summer results in larger contribution of oxidized organic vapors in new particle formation.The ammonium-excess condition in autumn promotes nucleation without the presence of organic compounds.

425
A particle growth mode that arises from condensation of semi-volatile organics from the bulk atmospheric composition is observed in F5.In particular, the substantial contribution of nitrate in F5 during summer could be seen as the possible promotion of organic nitrate formation in nitrate excess condition in higher radiation, condensing during the particle growth when the temperature is lower.F4 is seen as the continuation of previous growth modes with similar composition but larger particle size range.The factor with the largest particle diameter sub-range, F3, has a nitrate-rich composition across the seasons, 430 illustrating the ubiquitous importance of nitrate.
New particle formation is most pronounced with winds from the southwest-west, or north and east.These directions supply precursors gases, with the westerlies bringing SOx from the port of Rotterdam, southwesterlies bringing NH3 from agricultural emissions, and northerlies and easterlies bring organic vapors from the forest and nature areas.There is also an indication of 435 SOx sources from other urban centers to the north and east.The influence of the wind direction can be clearly seen during the summer, where instead of southern and western winds, the prevailing winds were from the north and east and brought abundant organics, resulting in the rapid growth of large amounts of OA.

Figure 1 .
Figure 1.Map of a part of the Netherlands showing the locations of the measurement stations Cabauw (main site) and Zegveld-Oude Meije (NH3 measurements).The province, sea, and neighbouring country names are indicated in italic and light grey.The big cities of Amsterdam, The Hague, Rotterdam, and Utrecht in the area collectively known as "Randstad" are situated in Noord-Holland, Zuid-Holland, and Utrecht provinces.The urban and harbor area of Rotterdam extends as Europoort-Maasvlakte to the mouth of the Maas River.100 https://doi.org/10.5194/egusphere-2023-554Preprint.Discussion started: 5 April 2023 c Author(s) 2023.CC BY 4.0 License.PMF analysis.Each organic fragment mass-to-ratio (m/z), species concentration, and size-binned particle concentration is treated as an individual variable in the PMF.The mass concentration variable values and errors were generated by Tofware v3.2.The particle size dataset from the SMPS instrument as categorized into 18 size bin variables

Figure 5 .
Figure 5.Time series of (a) particle size distribution (dN/dlogDp) in cm -3 with logarithmic scale in particle size obtained from SMPS measurements, (b) total mass loading calculated from ACSM species concentration (using Tofware) in µg m -3 , and (c) reconstructed PMF fraction (stacked) from analysis in May 2021.Orange outlined sections indicate periods during which high episodes of size-driven factors are observed.These episodes coincide with relatively low total aerosol mass conditions and high fine particle concentrations.Similar figures for June and September 2021 can be found in Fig. S3. 265 https://doi.org/10.5194/egusphere-2023-554Preprint.Discussion started: 5 April 2023 c Author(s) 2023.CC BY 4.0 License. 360

Figure 6 .
Figure 6.Normalized diurnal cycles in May 2021 of (a) the size-driven factors of F6 and F5, and SO2 as the sulfate precursor, and (b) the size-driven factors of F4 and F3.Notice that from F4 to F3, the factors' diurnal pattern shifted from high daytime to high nighttime concentration.Similar figures for June and September 2021 can be found in Fig. S5. 385

Figure 7 .
Figure 7. Timeseries of (a) wind direction (WD) color-coded with wind speed (WS), and (b) reconstructed PMF fractions F6 and F5 (stacked) corresponding to nucleation-mode and first growth-mode particles in May 2021.Orange outlined sections indicate high F6 and F5 episodes.We observe that new particle formation episodes are correlated with air masses from the southwesterly-westerly sector.Similar figures for June and September 2021 can be found in Fig.S6, showing more contributions from northerly-easterly wind directions.

Temperature ( o C) Downward shortwave radiation (W m -2 ) Precipitation (mm)
Wind rose plots for May, June, and September 2021.Winds from S up to SW sector were dominant in May.