A year-long seasonal sampling campaign was conducted during 2015/16 to collect size-resolved aerosol samples at the Lutjewad atmospheric measurement station located directly behind the dyke at the Wadden Sea in the Netherlands. The sampling equipment has been set up on a scaffold at about 10.5 m, which is higher than the dyke (see Fig.  in the Appendix). The sampling equipment consisted of a custom-made setup with two impactors (MSP-model 130-HFI) and a weather station.

The impactor setup is shown in Fig. . The two impactors are connected to the sampling pump via two controllable solenoid 3-way valves. At any given time either one of the impactors is sampling or the air is flowing through a bypass directly to the pump. The setting of the valves is controlled by a customised LabVIEW program, so that each impactor samples only from a given (custom-defined) wind sector. If air is coming from outside the wind sectors, neither impactor is sampling. Sampling will also only start when the wind velocity is above 2 m s⁻¹ to avoid baffling wind.

The location directly at the Wadden Sea made it possible to collect aerosol from a “land” sector (E–SSW) on one impactor and aerosol from the “sea” sector (WSW–NE) on the other. The land sector is mostly influenced by sources in continental Europe including possible transformation processes during long-range transport. This type of aerosol is of special interest, as it will be advected in the direction of the ocean and influence the radiative forcing over the North Sea . On the other hand the air from the sea sector is likely much cleaner and is mainly influenced by marine aerosol or regional sources, e.g. ship emissions from the highly frequented English Channel . In addition 72 h back-trajectories were calculated using the HYSPLIT trajectory model (see Figs.  and ) with a time resolution of one hour and an arrival altitude of 50 m a.g.l. (meter above ground level) for the sampling periods of each impactor.

The high flow cascade impactor had an inlet flow rate of 100 L min⁻¹ and consists of five stages and a back-up filter. It allows sampling in the following aerodynamic particle size ranges: 1.4–2.5, 0.77–1.4, 0.44–0.77, 0.25–0.44, 0–0.25 µm. To convert the measured concentrations to mass size distributions dM/dln⁡Dp, we assume a lower cut-off size of 0.05 µm for the back-up filter, as the contribution of particles below that size to the measured mass is negligible . In this study, quartz fibre filters (75 and 90 mm (back-up filter); Pallflex membrane filters, type Tissuquartz 2500QUAT-UP) were used as impaction substrates, which required customization of the distance ring for the last two impactor stages to compensate for the increased filter thickness compared to conventional aluminum foil. The quartz fibre filters were pre-cleaned at 650 °C for two hours and stored in the freezer at about -18 °C, before using them for sampling. After removing the filter samples from the impactor, they were wrapped in pre-cleaned (2 h, 550 °C) aluminum foils and stored in sealed polyethylene bags at about -18 °C in the freezer until analysis. Every part of equipment that came into contact with the samples, including impaction plates and tweezers etc., was first cleaned with acetone, then with ethanol and afterwards air-dried.

The filter sampling was conducted during each season of the year for at least one month, allowing for the collection of at least two complete sets of filter samples per season. A complete set of samples consists of one impactor filter for each particle size range from both the sea and the land wind sector. During the collection period the impactors were left in the sampling setup for approximately two weeks. If sampling took place with the specific sea-sector or land-sector impactor depended on the wind directions during this time period. Table  shows the collection periods (11–18 d) in the different seasons for the filter sets from this study. The sampling time per impactor varied between two and eleven days, depending on time periods with wind from a certain sector. At the end of each part of the sampling campaign, i.e. for each season, a complete set of impactor filter samples was left in the impactor in the sampling setup for approximately two weeks without the pump running. This filter set represents the sampling blank.

Filter samples for this measurement report were selected mainly according to the possibility of performing all the planned measurements, i.e. analyses of ions and levoglucosan, measurement of total carbon concentrations and isotopic analysis of organic carbon. Unfortunately only one sample from the sea wind sector could be analyzed, as the others did not contain enough particle mass. TC measurements and isotopic analyses were performed at the Centre for Isotope Research, University of Groningen in the Netherlands, while analyses of ions and levoglucosan were conducted at the Research Group Environmental Analytics, Institute of Chemical Technologies and Analytics, TU Wien in Austria. Depending on the TC mass of the impactor filter samples, aliquots were taken and sent to Vienna. Approximately one eighth of the filter sample area was cut for the land-sector filter sets and one quarter was taken for the sea-sector filter set. Each sample aliquot was separately wrapped in pre-cleaned aluminum foil and packed in sealed polyethylene bags. The parcel containing the aliquots was sent to Vienna by express post with an ice pack.

Due to the limitations in the available particle mass for this measurement study only a small number of filter sets could be analyzed for the different seasons of the year, see Table . The statements in the discussion section of this publication should therefore be interpreted as indications rather than as a comprehensive characterization of seasonal conditions or the study region.

Determination of anions (Cl⁻, NO2-, NO3-, SO42-), cations (Na⁺, NH4+, K⁺, Ca²⁺, Mg²⁺) and levoglucosan was performed with ion chromatography. For analyses of ions, sample aliquots were extracted with 3 mL ultrapure water and ultrasonic agitation, followed by centrifugation. Analyses of anions was done with a Dionex ICS1100 system (Thermo Fisher Scientific Inc.) using AS22 and AG22 columns, a self-regenerating suppressor and conductivity detection. The eluent was 4.5 mM Na₂CO₃ / 1.4 mM NaHCO₃, at a flow rate of 1 mL min⁻¹. Cations were analyzed with a Dionex Aquion system (Thermo Fisher Scientific Inc.) equipped with CS16 and CG16 columns, a self-regenerating suppressor and a conductivity detector. The eluent was 38 mM methanesulfonic acid at an eluent flow of 1 mL min⁻¹. Calibration was done with external standards, prepared from Merck stock solutions. Detection limits were at 0.01 µg mL⁻¹ for all ions. Levoglucosan was determined via anion-exchange chromatography and pulsed amperometric detection (HPAE-PAD) with an Dionex ICS3000 system using a CarboPacMA1 as an analytical and a guard column. Gradient elution was done with sodium hydroxide, using concentrations of 480–650 mM at a flow rate of 0.4 mL min⁻¹. Detection limit was at 0.015 µg mL⁻¹ for levoglucosan.

The total carbon concentration of the size-resolved aerosol filter samples was determined with a commercial thermo-optical OC-EC analyzer (Sunset Laboratory Inc., Model 5L). The samples were analyzed with a custom TC protocol, that combusts the sample in a mixture of helium and 2 % oxygen for 5.5 min at a temperature of 870 °C. The limit of detection for the measurement of carbon concentrations is given by the manufacturer of the OC-EC analyzer as 0.2 µg cm⁻². The measured total carbon concentrations have been corrected for the instrument blank and the sampling blank. The instrument blank was measured daily and sample concentrations were corrected using the instrument blank determined on the corresponding measurement day. The instrument blank for this study was on average 0.27 µg cm⁻² with a standard deviation of 0.15 µg cm⁻² (n=58). The sampling blanks were analyzed with the EUSAAR_2 protocol and contained only organic carbon with an average concentration of 0.69 µg cm⁻² with a standard deviation of 0.24 µg cm⁻² (n=18).

OC/EC analysis of the impactor filter samples was not possible due to the inhomogeneous deposits on the impactor samples.

The stable carbon isotope ratio R(13C/12C) of OC in a sample is given in the delta notation, which is defined as follows: 1δ13C/12C,sample/reference=R13C/12C,sampleR13C/12C,reference-1 The reference is the international Vienna Pee Dee Belemnite (VPDB) standard and delta values are expressed in ‰. In the following text δ(¹³C) is used as the short form of δ(13C/12C, sample / reference).

A detailed description of the setup and the measurement procedure is given in . Briefly, measurements of δ(¹³C) values are conducted with a system that combines a OC-EC analyzer (Sunset Laboratory Inc., Model 5L), where OC is converted to CO₂, with a continuous flow isotope ratio mass spectrometer (IRMS) (Micromass now Isoprime, model Optima). The interface between these two instruments serves to collect, purify and concentrate the extracted CO₂ from the aerosol filter samples. In the Sunset oven the organic carbon is desorbed in helium carrier gas at three different temperature steps (200, 350, 650 °C). The manganese dioxide (MnO₂) catalyst in the Sunset OC-EC analyzer ensures the complete oxidation of the carbon released during the different desorption steps to CO₂.

Thermal desorption for the δ(¹³C) analysis is performed in an inert helium atmosphere, where OC charring can occur and the resulting charred carbon is not included in the subsequent isotope analysis. Previous studies have estimated that this does not strongly alter the δ(¹³C) signature .

For the analysis of the coastal filter samples, the system described by was updated to include a reduction oven, which is installed downstream of the Sunset OC-EC analyzer. This oven contains copper and silver to remove traces of oxygen, nitrogen oxides and halogens at a temperature of 600 °C. This treatment was found to significantly improve the repeatability of the filter samples from this coastal location, but is usually not necessary for filters from other locations.

δ(¹³C) values were calculated with respect to two reference gas peaks, measured before and after the sample peak. If these two values differ by more than 1 ‰, this certain measurement is rejected from further analysis. δ(¹³C) values calculated with respect to the first reference peak are presented in this study. δ(¹³C) signatures are calibrated using a two-point linear scale correction based on two different reference materials, which are measured every day. The linear scale correction uses a weekly average of these reference materials and is applied to all samples measured in the respective week. These two in-house reference materials are both caffeines and one of them is enriched in ¹³C. Their nominal δ(¹³C) values are -38.2 ‰ and 0.61 ‰. A third international reference material is analyzed every day and is used for quality control. This internationally agreed upon secondary reference material is L-Valine and has a reference value of -24.03±0.04 ‰, which is in the range of the expected δ(¹³C) values of aerosol particle samples. The average δ(¹³C) value of the measured L-Valine reference for a time span of about two years is -24.07 ‰ and the standard deviation amounts to 0.22 ‰. This is an indicator for the precision and the uncertainty of the δ(¹³C) values presented in this work.

The radiocarbon isotopic composition of an aerosol filter sample is reported as fraction modern (F(¹⁴C)) according to the nomenclature established by : 2F14C/12C,sample/OxII=R14C/12C,sample0.7459⋅R14C/12C,OxII Hereby the 14C/12C ratio of the sample is given in relation to the ratio of the standard (oxalic acid, OxII). The factor 0.7459 converts the 14C/12C ratio of OxII to the value defined by international agreement as representative for an unperturbed atmosphere in the year 1950. The 14C/12C ratio of both the sample and the standard are normalized to a δ(¹³C) of -25 ‰. In the following text F(¹⁴C) is used as the short form of F(14C/12C, sample / OxII).

In this study F(¹⁴C) of organic carbon is presented. The combustion, extraction and conversion to CO₂ was done using the Automated Aerosol Combustion (AAC) system described in . Briefly, the aerosol filter sample was combusted in pure oxygen at 375 °C for 8 min. The separation of CO₂ from other interfering gases (e.g. SO₂, NO_x, H₂O or halogens) is done with the help of a reduction oven containing copper and silver, and two water traps. Until analysis the CO₂ is stored in evacuated flame sealed glass tubes.

For the F(¹⁴C) analysis, extraction at 375 °C in pure oxygen extracts approximately 80 % of the organic carbon as defined by thermal-optical analysis, with the most refractory OC omitted due to potential overlap with EC. However, the use of oxygen largely prevents charring, and the less refractory carbon fractions are extracted with an efficiency close to 100 %.

For radiocarbon measurement the CO₂ is transferred via the gas inlet system to the ion source of the Mini Carbon Dating System (MICADAS). This accelerator mass spectrometer is described in more detail in . The F(¹⁴C) values are corrected for the memory effect, the instrument background is subtracted and finally all values are normalized to the average of the oxalic acid (OxII) gaseous standard, which are analyzed in every batch . Furthermore secondary standards are analyzed alongside the unknown samples, which serve as a control of the measurement and data correction .

To correct for the contamination introduced during combustion in the AAC, small amounts of standard material undergo the same process of combustion and radiocarbon measurement. Here OxII serves for the determination of the fossil contamination and a ¹⁴C free background wood standard serves to determine the modern contamination.

Figure  shows an overview of the mass size distributions dM/dln⁡Dp of all analyzed chemical constituents: inorganic ions, levoglucosan and total carbon. The different icons indicate a specific set of impactor filters and the different colors characterize a specific season of the year (winter – blue, spring – green, summer – red, autumn – gray; exception: yellow – summer, sample type “sea”; also consider Table ).

Figure  highlights the relative contribution of the main constituents (NH4+, NO3-, SO42-, TC, Na⁺, Cl⁻) for four typical impactor samples. Figure a and b compare in more detail the mass fractions of a sample collected in spring (SP16_L1) to a sample collected in autumn (AU15_L2) to highlight seasonal differences. Figure c and d show two filter sets collected in summer during the same time period; one set from the land (SU16_L4) and the other from the sea wind sector (SU16_S4). In addition the 72 h back-trajectories (HYSPLIT trajectory model, time resolution one hour, arrival altitude 50 m a.g.l.) are shown for each filter set.

As observed in earlier studies in the Netherlands the largest contribution to the aerosol particle mass comes from nitrate, ammonium and total carbon. The outflow aerosol from the land to the sea is dominated by ammonium nitrate as the main water soluble species. A main reason for the large influence of ammonium nitrate is the combination of intensive agriculture, responsible for high ammonia emissions, and the high traffic and industrialization, which emit nitrogen oxides. Most inorganic components show the highest mass concentrations at a size range between 0.5 and 1 µm, indicative of aged aerosol. This shows that the measurement site is influenced mainly by long range transport and not so much by local sources. The size distribution changes with air mass origin and residence time in the atmosphere, as can be seen comparing the filter set from autumn with the set from summer. For the autumn sample, the back-trajectories (see Fig. ) show a long residence time over continental Europe and the mode of the mass distribution is at large particle diameters of around 1 µm. The summer sample on the contrary mostly has a much shorter residence time over land close to the sampling site and the mode of the size distribution is at smaller diameters of around 0.6 µm.

Particulate ammonium is the main cation to neutralize the most important atmospheric anions to form ammonium nitrate, ammonium sulfate and ammonium chloride. Formation of ammonium nitrate requires high concentrations of NH₃, because the more stable ammonium sulfate is preferentially produced. Ammonium nitrate is semi-volatile and consequently the mass concentration of particulate nitrate depends on temperature, with lower concentration in summer compared to the colder periods of the year. Especially high particulate nitrate concentrations occur in samples collected at the end of the winter period (WI16_L2) and in spring samples with maximum concentrations between 0.5 and 1 µm (WI16_L2: (0.6 µm, 4.3 µg m⁻³), (1.1 µm, 3.0 µg m⁻³); SP16_L1: (0.6 µm, 4.6 µg m⁻³), (1.1 µm, 3.6 µg m⁻³)). The highest mass concentrations of ammonium can be found in spring samples for particle sizes ranging from 0.44 to 1.4 µm (SP16_L1: (0.6 µm, 1.75 µg m⁻³), (1.1 µm, 1.4 µg m⁻³)), related to fertilizer use. The lowest concentrations are observed for samples from the sea-influenced sector (0.01–0.2 µg m⁻³), with the lowest relative contribution to the mass of the particles from the highest size range. Otherwise, differences in concentrations between samples collected during different seasons are not very strong. Currently, it is up for debate if reducing emissions of NH₃ or NO_x are more beneficial for reducing particulate matter . However, our data show especially high contributions of ammonium nitrate to particulate matter, when ammonia emissions through fertilizer use are high in spring. Since NO_x emissions are much more constant over the year, this indicates that low ammonia availability is at least partially a limiting factor for ammonium nitrate formation.

The particle mass concentration of sulfate does not show any distinct seasonal pattern among the analyzed samples. In Fig.  the total mass of sulfate is shown. The proportion of non-sea salt sulfate was calculated with sodium as a reference for sea salt and was found to be more than 97 % for all particle sizes except the largest size range for the land-sector filter set. Aerosol particles from the largest size range (1.4–2.5 µm) show a clear impact from sea salt emission sources with a proportion of up to 21 % sulfate from sea salt and mainly for filter sets (e.g. in Fig. c) under the influence of regional and coastal air masses. The sea-sector filter set has a proportion of 7 % sulfate mass from sea salt in the size range of 0.77 to 1.4 µm and 46 % in the size range of 1.4 to 2.5 µm. For the remaining particle size ranges of the sea-sector filter set the non-sea salt sulfate mass represents more than 98 % of total sulfate. Furthermore for the sea-sector filter set, sulfate is the main component next to TC for all sizes besides the largest collected size range. For sulfate aerosols, the ocean itself is a source with e.g. a total concentration of non-sea salt SO42- of around 43 ng m⁻³ on average for a remote station on the Svalbard Islands . The sea-sector filter shows much higher SO42- concentrations than this (0.1–0.5 µg m⁻³), suggesting that the combustion of sulfur-containing fossil fuels by ship engines, as an important emission source of the precursor gas SO₂ , may contribute to the observed sulfate concentrations. Since TC represents the second-largest fraction of the sub-micron aerosol mass in the sea-sector filter, this is consistent with a contribution from ship emissions to the observed aerosol at this coastal location. However, TC is not specific to shipping emissions, and the back-trajectories of the sea-sector filters (see Fig. d) show that part of the air masses passed over Great Britain, indicating a possible influence by other sulfate sources, such as vehicular traffic emissions. The relative mass contribution of nitrate in the sea-sector filter samples is negligible for particles smaller than 0.44 µm (see Fig. d). This may indicate a stronger influence of ship emissions, as the particle mass from ship emissions is generally dominated by sulfate and organic matter . In contrast, measurements of the London plume showed a substantial nitrate contribution of 24 % to sub-micron aerosol, comparable to the contributions of sulfate (24 %) and organic aerosol (29 %). The absence of a similarly strong nitrate contribution in the sub-micron particles of the sea-sector filter samples therefore supports the interpretation that shipping emissions may represent an important source of the observed aerosol. However, the low nitrate contribution and high sulfate and TC concentrations in the sea-sector filter set are consistent with an influence from shipping emissions but do not allow unique source attribution. In general, sulfate is the only major species (except for NaCl), where the absolute mass concentrations of the sea-sector filter is within the range of the land-sector filters and even exceeds the land-sector filters (for size range around 0.35 µm sea-sector filter: 0.4 µg m⁻³, land-sector filter: 0.16–0.3 µg m⁻³). Therefore most likely ship emissions may represent also a very relevant source not only locally, but for coastal locations in the Netherlands in general. Nonetheless the absolute concentrations of ammonium sulfate are much lower than those of TC and ammonium nitrate, which is most likely related to the substantial reductions in SO₂ emissions from major sources, including the industrial sector, following the implementation of emission-control measures and regulations .

The total carbon mass shows higher concentrations in the winter and autumn samples, which is most likely due to domestic biomass burning with the starting cold season. Potassium and levoglucosan, typical biomass burning tracers, also show clearly higher values in autumn and winter samples and lower values in summer samples and for the sea-sector filter set. Local and regional sources seem to make a significant contribution, because these tracers are especially enriched in the lower size ranges compared to the other analyzed components. The contribution of carbon from traffic emissions is assumed to be constant throughout the seasons. Local traffic emissions would cause elevated TC mass concentrations in the lower size ranges, as the combustion of fuel produces small aerosol particles. Our samples do not show this influence, which underlines the rural character of the measurement site. The sea-sector filter set, which was collected in summer, shows the lowest TC mass concentrations overall (0.1–0.34 µg m⁻³), but for the smallest size range TC concentrations are comparable to other samples (0.05–0.23 µg m⁻³). The source of TC cannot be uniquely identified, but shipping emissions may be one possible contributor. In general, one would expect for a rural station higher TC concentrations in summer than in winter or autumn as biogenic secondary organic material contributes strongly to the aerosol mass. The measured concentrations do not show this and it was also not observed in earlier studies in the Netherlands . The main reason here is probably the higher planetary boundary layer in summer than in the colder seasons. However, in the smallest two size classes, the mass concentrations of TC in the summer samples are comparable to or even higher than those in the autumn and winter samples. This indicates, that biogenic SOA, produced regionally in the Netherlands during the summer, probably contributes more strongly to smaller particle sizes. This is supported by air mass back trajectories, which during the summer show air mass origin over the Atlantic and transport over the Benelux and Northern France resulting in regional pollution advected to the site. In autumn, the air is coming from central Europe representing integrated European outflow and much longer aging times.

Sea salt is a source of the following analyzed ions, stated from the highest to the lowest proportion by weight appearing in sea salt: chloride (Cl⁻), sodium (Na⁺), sulfate (SO42-), magnesium (Mg²⁺), calcium (Ca²⁺), potassium (K⁺) . Particles with diameters >1 µm account for most of the PM_2.5 sea salt aerosol mass, which is reflected in the aerosol mass size distributions shown in Fig. . Therefore, even though sea salt can significantly contribute to PM_2.5, the contributions to the sub-micron range are very low. Obviously the sea-sector filter set shows the highest chloride, sodium and magnesium mass concentrations for the highest size range, where the sodium chloride mass accounts for 40 % of the total constituents (Fig. d). Even the land-sector filter set from the same time period (SU16_L4) shows a significant contribution from sea salt for the highest size range, which is related to air masses originating over the sea before crossing land as shown by the corresponding trajectories.

Despite the fact that chloride takes up the largest mass proportion of the sea salt constituents, its mass concentration or rather the concentration in mol m⁻³ (see Fig. ) is lower than sodium for the largest size range in the coastal aerosol samples. The difference is even clearer for the impactor stages with mean diameters of around 0.6 and 1 µm (see Fig. ), where sodium shows elevated values in the sea-sector filter compared to the other samples, but chloride does not. This is consistent with the ageing of sea spray through reaction with nitric acid (HNO₃), forming gas-phase hydrogen chloride (HCl) and particulate-phase sodium nitrate (NaNO₃) . Given the proximity of the sampling site to the coast, this reaction may occur on relatively short timescales after the air mass makes contact with the coast. Laboratory studies have demonstrated rapid HNO₃ uptake and chloride depletion in deliquesced NaCl particles under atmospherically relevant conditions . For the largest size range, the chloride is much less depleted than for the smaller sizes.

In the middle size ranges (mean diameters of around 0.35 and 1 µm), chloride mass concentrations (cf. concentrations in mol m⁻³ in Fig. ) are higher than sodium mass concentrations for the land-sector filters, especially in samples collected in winter and autumn. Chloride is then likely present in the form of ammonium chloride (NH₄Cl), a semi-volatile compound, which explains higher concentrations in the samples collected during the colder seasons. Magnesium shows a very similar size distribution as sodium with the largest values for the sea-sector filter set and at larger particle sizes, indicating that its main source is sea spray.

Calcium has a very distinct mass size distribution with elevated values for the lowest and the highest size range. An emission source for calcium is sea salt, but this has only a minor contribution. Another important source is mineral crust material, especially limestone , which produces primary coarse mode particles by wind erosion. In general dry conditions and high wind velocities cause more mineral crustal material to become airborne. This explains the high particle mass concentrations for the largest size range and increased values for winter and spring samples for this size range. Elevated values for the lowest size range could occur, because of particle bounce in the impactor , which causes large particle not to be collected on their size corresponding impaction plate, but to bounce off and eventually end up on the back-up filter (lowest size range). This happens especially with dry particles, such as mineral dust. However, particle bounce is less likely on quartz fibre filter compared to smooth substrates. The reason for elevated particle mass concentrations of calcium in the lowest size range remains unclear.

Figure  shows the fraction modern of organic carbon (F(¹⁴C, OC)) as a function of particle size for the analyzed aerosol filter samples. An F(¹⁴C) of 0 % corresponds to pure fossil origin of the OA, whereas F(¹⁴C) of around 100 % indicates biogenic origin. The particle size is shown as the geometric mean of the lower and the upper cut-off-size of the relevant impactor stage. If two stages were combined in order to have enough material for the analysis of the 14C/12C ratio (i.e. for some spring and summer samples), the geometric mean of the lower cut-off diameter for the smaller stage and the upper cut-off for the larger stage is used. Some of the autumn filter samples have been analyzed twice and the values show a high repeatability for F(¹⁴C, OC).

The organic carbon of all filter samples is clearly dominated by modern sources, as the lowest F(¹⁴C, OC) value is about 67 %. The highest F(¹⁴C, OC) occurs in spring samples with a value of about 103 % in the size range that covers particle sizes from 0.44 to 0.77 µm, which is very close to the F(¹⁴C) of atmospheric carbon dioxide of approximately 104 % . In general, the impactor samples taken in spring show elevated F(¹⁴C, OC) compared to samples from the other seasons. For samples from all seasons F(¹⁴C, OC) is lowest at small particle sizes and increases towards the particle size range of 0.44 to 0.77 µm. Usually, a slight decrease for the larger particles can be observed. Even for this relatively remote coastal site in the Netherlands quite far from major roads and urban centres a clear influence of fossil sources on the smaller particles is evident. These particles most likely originate from regional vehicular and/or ship traffic. For the winter samples this trend is clearest. F(¹⁴C, OC) is 68 % for the smallest size range, increases to 84 % for the middle size range and decreases to 70 % for the largest size range. To our knowledge, this is the first time that this significant size dependence in the fossil contribution to OC has been demonstrated in Western Europe. In contrast a study in a megacity in China showed no significant size dependence in fossil contribution to OC .

The main sources for contemporary carbon are biogenic SOA and biomass burning, but previous studies show, that also primary biogenic particles, e.g. pollen fragments, or cooking emissions can contribute to high values of the fraction modern. In this remote area cooking is not a very likely source as there are only small villages within 20 km of the site.

Higher values of F(¹⁴C, OC) in spring compared to all other seasons were also observed in another seasonal study of radiocarbon in the Netherlands . In this study, F(¹⁴C, OC) was analyzed for PM2.5 without any size dependence. Average values of F(¹⁴C, OC) for summer (71 %), autumn (78 %) and winter (74 %) varied within a narrow range of about ±4 %, but the average F(¹⁴C, OC) in spring was much higher with a value of 89 %. This is roughly comparable to the weighted average over the whole size range in our study (summer 75 %, autumn 78 %, winter 76 %, spring 90 %). discuss, that elevated values of F(¹⁴C, OC) in spring are most likely caused by biogenic SOA and prevalent during the growing season. In the highly agricultural region of the Netherlands maximum biogenic SOA production thus seems to occur in spring and not in summer, such as in other, more forested, regions.

The reason for the elevated F(¹⁴C, OC) values in spring remains unclear. Enhanced secondary organic aerosol formation associated with agricultural activities, increased biogenic precursor emissions, or contributions from primary biological particles such as pollen fragments are all possible explanations. However, the available dataset does not allow us to distinguish between these possible mechanisms. Further targeted investigations are therefore needed to better understand the processes responsible for the elevated F(¹⁴C, OC) values observed in the spring samples.

F(¹⁴C) of SOA depends on the origin of the precursor gases that are involved in the formation of secondary organic carbon. If these gases primarily originate from fossil sources, then the formed particulate matter will have low F(¹⁴C) values. In our case an influence from modern sources on SOA is more prominent, as F(¹⁴C, OC) is above 65 % for all size ranges and sources surrounding the sampling station are predominantly biogenic. Even though biogenic sources and SOA production are mainly enhanced in warmer seasons, F(¹⁴C, OC) in winter samples shows similar values to summer samples. This is likely caused by biomass burning as an important source for OC in winter and autumn, which increases F(¹⁴C, OC). Biomass burning in the Netherlands mostly occurs in residential heating with fire wood or pellets. This type of heating is more common in rural areas than in cities.

Figure  shows the δ(¹³C) values of organic carbon (δ(¹³C, OC)) as a function of particle size for different desorption temperatures (200, 350 650 °C). The color coding for the samples from different seasons of the year is the same as used in Fig.  (winter – blue, spring – green, summer – red, autumn – gray; exception: yellow – summer, sample type “sea”; see Table ). The shaded horizontal areas show the range of typical δ(¹³C, OC) values for particles emitted from primary sources, as expected for this region . Non fossil sources, namely biomass burning, are shown in green and traffic emissions are shown in gray. For biomass burning aerosol the range encompasses emissions from burning of alder, oak, birch, ash and Quercus ilex (oak) wood. These biomass burning sources represent characteristic material for heating, that can be bought, e.g. in the form of pellets, on the Dutch market. Traffic emissions were measured on filters collected in a tunnel of a main city traffic street in Naples and directly collected from a city bus exhaust . These samples are selected as they represent a typical mixture of traffic emissions, that can be expected in Western Europe. Unfortunately not so many source reference measurements are published so far for this type of analysis with the different temperature steps, especially focused on the Western European region.

For some impactor filter sets δ(¹³C, OC) values were not available for every particle size, as e.g. the measurement did not satisfy the quality control or not enough filter material was left after chemical analysis and the analysis of the 14C/12C ratio. On that account for some seasons, especially winter, it is not possible to draw a comprehensive conclusion for the trends of the δ(¹³C, OC) values over the whole particle size range. The measurement of the autumn filter set (AU15_L2) was repeated for some size ranges. Most of these δ(¹³C, OC) values are repeatable within the measurement uncertainty.

An overall increase of δ(¹³C, OC) can be observed with increasing desorption temperature. This was also observed in previous studies on ambient aerosol . Since the same increase is typically not necessarily observed in source samples , we conclude that it is related either to differences in contributing sources at the investigated desorption temperatures or to different atmospheric transformation processes that influence the particular OC fractions.

Considering the particle size distribution, the δ(¹³C, OC) values tend to increase with increasing particle size. This increase in δ(¹³C, OC) with increasing particle size is more pronounced for higher desorption temperatures and less evident at 200 °C. At higher temperatures many δ(¹³C, OC) values, especially for the size range 0.44 to 0.77 µm and larger, are higher than the δ(¹³C, OC) values of the possible primary sources. The filter sets from summer and autumn show a clearer increase of δ(¹³C, OC) with increasing particle size, especially for 350 °C and 650 °C. found an increasing mass fraction of OC at 650 °C for larger particles and also higher δ(¹³C, OC) (650 °C) values with increasing particle size. The observation was explained with photochemical aging of the aerosol, as it especially occurred in combination with low and moderate PM_2.5 pollution events, where regional aged aerosol contributes significantly to urban pollution. At 200 °C most δ(¹³C, OC) values are lower compared to the primary sources (horizontal green and gray areas). This is especially the case for filters from spring and summer, where δ(¹³C, OC) values are all lower than the values of the primary sources. Lower values of δ(¹³C, OC) for the OC fraction desorbed at 200 °C were also found for samples from the Tibetian Plateau , different sampling areas in Lithuania and samples collected in Beijing .

Figure  compares calculated δ(¹³C, OC, prim) (black squares) based on a hypothetical mixture of only primary sources (in this case, traffic emissions and biomass burning) to the measured δ(¹³C, OC) (red circles) for autumn. The hypothetical primary source mix is based on the F(¹⁴C, OC) values from Fig. , which give the relative contribution of fossil and modern sources to OC. For example a F(¹⁴C, OC) value of 80 % means, that roughly 80 % of OC originated from modern sources and 20 % originated from fossil sources. The δ(¹³C, OC) values assigned to fossil and modern primary sources are represented by the values for vehicular sources and biomass burning, respectively, as shown in Fig.  by the shaded areas. For the calculation the respective δ(¹³C, OC) values of the sources are averaged and are depicted as horizontal lines in Fig.  for every temperature step. Modern sources, represented here by biomass burning, are shown as green line and fossil sources, represented here by traffic sources, are shown as gray line. The calculation of the theoretical δ(¹³C, OC, prim) is summarized in the following equation: 3δ13C,OC,prim=F14C,OC,sample104%⋅δ13C,OC,modern+1-F14C,OC,sample104%⋅δ13C,OC,fossil 104 % is the F(¹⁴C) value of atmospheric carbon dioxide . Equation () provides an illustrative two-endmember comparison between modern and fossil OC and should not be interpreted as a unique quantitative source apportionment. Modern OC may originate from sources other than biomass burning, while fossil OC is not necessarily attributable to only vehicular traffic.

The most obvious observation from Fig.  is the low variability (< 1 ‰) of the calculated δ(¹³C, OC, prim) with changing particle size compared to values of the real sample. The calculated δ(¹³C, OC, prim) do not vary strongly with the temperature step or particle size. This is also in line with findings of a study investigating the δ(¹³C) of OC in primary emissions , where δ(¹³C, OC) did not vary much with changing desorption temperature. Basically, if the organic aerosol were dominated by a primary source mix, without the contribution of SOA or aging processes, we would expect δ(¹³C, OC) to vary only in a relatively narrow range.

On the contrary, the δ(¹³C, OC) of the sampled aerosol particles show an increase with increasing particle size. This means smaller, on average less aged particles have lower δ(¹³C) values than larger particles. With respect to desorption temperature, at 200 °C all values of δ(¹³C, OC) of the samples are depleted compared to the calculated values. For the higher temperatures (350 and 650 °C) the measured δ(¹³C, OC) values are, except for the smallest particle size range, enriched compared to the theoretical δ(¹³C, OC, prim).

The depletion at 200 °C is most likely caused by the influence of recently formed SOA as it tends to desorb at lower temperatures . Due to the kinetic isotope effect, ¹²C is preferred during the formation of SOA from the precursor gas and accumulates in the reaction products that condense on the aerosol. As a consequence the 13C/12C ratio in the formed SOA will be less than in the precursor gas and likely less than in the primary organic particle matter. If incomplete, this process leads to a depletion of ¹³C (i. e. low δ(¹³C) values) in the particle phase of the aerosol. The original primary particle, not influenced by atmospheric transformation processes, is in our case represented by the theoretical δ(¹³C, OC, prim). SOA condenses on all particles, irrespective of their size. However, proportionally more SOA is added to the smaller particles. The lower δ(¹³C, OC) values of the smaller particles observed in this study are consistent with this fact. Since SOA can also contain low- and non-volatile compounds, a similar effect and size dependence can also be seen for higher desorption temperatures. However, for higher desorption temperatures only the smallest particles are depleted in ¹³C compared to the estimated primary signature. As shown in Fig. , δ(¹³C, OC) is especially depleted in the spring and summer samples, compared to samples from the other seasons and the primary source values. This is consistent with enhanced SOA formation in these warmer seasons.

The enrichment of the measured δ(¹³C, OC) compared to δ(¹³C, OC, prim), especially at 650 °C, is most likely caused by aerosol aging. Photochemical aging causes a net loss of organic carbon from the particle and the isotopically lighter carbon ¹²C is preferentially lost. If the particle is only partially oxidized, the remaining carbon in the particle is enriched in ¹³C compared to carbon of the original particle. As reflected by the δ(¹³C, OC) of the samples (see also Fig. ), aerosol aging is more relevant for larger particles (accumulation mode) as these particles have a longer residence time in the atmosphere and therefore are involved in long range transport. The smaller particles (see Fig. ) show δ(¹³C, OC) close to the primary emission sources at higher desorption temperatures. The increase of the measured δ(¹³C, OC) values with particle size is much stronger at 650 °C than for lower temperatures. Especially for this filter set from autumn the influence of aged regional aerosol is larger than the effect of SOA production.