The sampling site is located at the air quality station operated by the Interuniversity Laboratory of Atmospheric Systems (LISA), which is inside Université Paris-Est Créteil (UPEC) in the south-east of the Paris agglomeration (48.79∘ N, 2.44 ∘ E) (Fig. 1). The study site is in close proximity to various sources of pollution including nearby industries as well as an incinerator, highways, railway stations, and construction sites. Between July 2021 and July 2022, daily rainfall depths measured using a Précis-Mécanique rain gauge (model 3070 A) (0.2 mm precision) at the study site ranged from 0.2 to 37.6 mm. Nineteen percent of rainy days presented rainfall depths lower than 0.4 mm, 12 % were between 0.4 and 1 mm, 40 % were between 1 and 5 mm and 13 % were higher than 10 mm. The sampling strategy is to investigate case studies sampled during an intensive measurement campaign during the winter and spring of 2022. During this period, the daily average PM10 (PM2.5) concentrations were around 17.5 (11.2) µg m-3 with values reaching up to 57.5 (43.0) µg m-3. Wet deposition collection is performed with a sequential sampler specifically developed at the LISA (Fig. 1a). Sampling, conditioning and analysis of rain samples are described in detail in Audoux et al. (2023); thus, it is only briefly reminded here.

The rain is collected using a Teflon pyramid funnel with a collection surface of 1 m2 in combination with a sampling unit. This unit enables the automatic collection of up to 24 consecutive fractions of rain, adjustable from 0.05 to 2.0 mm, to study dissolved and particulate phase of the wet deposition. The sampling is conducted based on volume, and as a result, it is dependent on the rainfall rate. The sequential sampler is able to correctly collect rainfall fractions for low rainfall intensities, as well as for more intense rainfall recorded by the rain gauge and disdrometer, in comparison to standardized measurements (Audoux et al., 2023). The materials that make up the sampler have been chosen to allow for analysis of the ionic and elemental composition of the dissolved and particulate phase at high and low concentration levels (from several milligrams per litre (mg L-1) to hundreds of nanograms per litre (ng L-1)). The sampling bottles and materials that came in contact with the samples underwent a thorough washing protocol in a clean-room laboratory with ISO-7 and ISO-5 level controls.

A summer rain event was collected in July 2021 (R1), and a winter rain event were collected in February 2022 (R2) (Audoux et al., 2023). These case studies are completed here with six additional events collected in late winter and spring 2022, between March (R3, R4, R5, R6 and R7 rain events) and April (R8 rain event). For the 8 rain events studied, between 11 and 32 consecutive fractions have been sampled, with the latter being collected within 10 s to 2 h, depending on the rainfall rate.

Concomitant measurements on atmospheric aerosols and meteorological parameters during the rain sampling are important for a more in-depth understanding of wet deposition mechanisms. Therefore, PM2.5 and PM10 aerosol mass concentration, as well as the particle size distribution (PSD) between 0.18 and 18 µm, were measured using FIDAS (Fig. 1d), equipped with a TSP Sigma-2 inlet, with a 1 min time step. The FIDAS is an instrument used for regulatory air quality measurement of PM2.5 and PM10 mass concentration (LCSQA, 2021). Moreover, PM10 aerosol particles are sampled on polycarbonate membranes (Nuclepore^®, 0.4 µm porosity) using a PM10 head sampling (Fig. 1c). Air sampling is done between 15 and 24 h before the start of the rain and is stopped at the beginning of the rain event, within 1 min after removing the cover from the sequential sampler, while the first fraction is being collected. This allows for characterizing the chemical composition of the atmospheric aerosol prior to rainfall. Rainfall rate and droplet size distribution (DSD) are measured using an OTT PARSIVEL^® (PARticle SIze and VELocity, Fig. 1e, the Supplement Fig. S1) optical disdrometer with a time resolution of 30 s. In parallel, wind direction and wind speed as well as air temperature and relative humidity are measured using instrumentation from Campbell Scientific© and are recorded with a time step of 1 min. The cloud base height and homogeneity of the atmospheric column are measured using a ceilometer (Vaisala CL31, Fig. 1b and S2). PARSIVEL disdrometers and ceilometers are typically used in multiple measurement networks for precipitation and cloud base height characterization (e.g. Haeffelin et al., 2005; Tapiador et al., 2010). FIDAS, ceilometer and disdrometer measurements are made continuously at the study site, while aerosol filter sampling and deposition measurements are made on alert before or during rain events. This makes it necessary to regularly follow-up the precipitation alerts.

After sampling, rain samples are quickly processed for ionic and elemental analysis, usually within a time frame of 1 to 12 h after the end of rainfall. If immediate processing is not feasible, the samples are kept in a cool and dark environment at 6 ∘C and are processed within 24 to 48 h. Treatment, filtration and conditioning are done in a clean-room laboratory with ISO-6 level controls, under a laminar-flow hood (U15 filter), which is estimated to be equivalent to ISO 3. A pH meter (Mettler Toledo Seven2Go) is used to measure the pH of each sequential sample. Samples are then filtered through pre-cleaned Nuclepore^® polycarbonate membranes with a porosity of 0.2 µm to separate the particulate phase from the dissolved phase. Following Audoux et al. (2023), the dissolved phase is then divided into two fractions. The first fraction (10 mL aliquot) is frozen until the analysis of water-soluble major inorganic cations (Na+, NH4+, K+, Mg2+, Ca2+), anions (Cl-, NO3-, PO43-, SO42-) and organic ions (HCOO-, CH3COO-, C2H5COO-, CH4O3S, C2O42-) by ion chromatography (Compact IC Flex, Metrohm^®, PRAMMICS platform). The second fraction (two 15 mL aliquots) is acidified to pH = 1 with nitric acid (Suprapur^®) and stored at 6 ∘C until analysis of water-soluble Al, Ba, Ca, Cr, Fe, K, Mg, Mn, Na, Ni, P, S, Si, Ti and Zn by an inductively coupled plasma atomic emission spectrometer (ICP-AES, Spectro ARCOS Ametek^®). The membranes are dried under a laminar-flow hood and conditioned for 48 h at a constant relative humidity of 45 %–50 % and at T= 20 ∘C prior to weighing using a precision microbalance (Mettler Toledo^® XPR26C, PRAMMICS platform). In order to accumulate a sufficient amount of material for analysis, several rain sequential samples can be filtered through the same filter. Conversely, when the particulate load is too high, rain fractions can be filtered through multiple membranes. Elemental composition (Al, Ba, Ca, Cr, Fe, K, Mg, Mn, Na, Ni, P, S, Si, Ti and Zn) of the particulate phase is determined using X-ray fluorescence spectrometer (XRF, ZETIUM 4 kW, Malvern Panalytical, PRAMMICS platform). The 0.4 µm porosity Nuclepore^® membranes are also analysed using XRF to characterize the elemental composition of the aerosol in the air prior to rainfall events. Our strategy is to monitor the elemental inorganic fraction of the aerosol in order to link it to the rainfall composition. This therefore allows us to characterize about 40 % of the average aerosol composition in the Paris region (Airparif, 2021).

The origin of scavenged aerosol particles can be discussed in relation to their chemical compositions and the trajectory of the air masses. We calculated enrichment factors (EFs; Taylor and McLennan, 1985) in order to determine the origin of elements found in the rain samples. Al is used as the reference for the Earth's crust (hereafter referred to as EFXcrust), and Na is used as the reference for sea salt (hereafter referred to as EFXseasalt). Equation (1) is used to calculate EF as follows: 1EFX(%)=([X]/[ref])rain([X]/[ref])crustorseasalt×100, where ([X]/[ref])rain correspond to the ratio between the element X and the reference (Al or Na) concentrations in rainwater samples and ([X]/[ref])crustorseasalt the concentrations in the continental crust or in the sea.

To complement local wind measurements at the study site, air mass trajectories were calculated using the HYSPLIT model (https://ready.arl.noaa.gov/HYSPLIT.php) (Draxler and Rolph, 2012). HYSPLIT is a retro-trajectory analysis used to study local-to-continental air mass dispersion and transport of atmospheric compounds (Celle-Jeanton et al., 2009; Bertrand et al., 2008; Calvo et al., 2012) and to determine the origin of air masses to identify sources of atmospheric substances, e.g. mineral-dust, sea salt or anthropogenic sources (Vincent et al., 2016; Anil et al., 2017). Here, 48 or 120 h, depending on the event, backward trajectories were computed by the HYSPLIT model with the Global Forecast System (GFS) (0.25∘, global) from the study site (47.79∘ N, 2.44∘ W) at the surface (0 m a.g.l.) and at the cloud base height measured by the ceilometer.

Eight rainfall events constitute a data set illustrating various cases in terms of aerosol concentrations and compositions as well as precipitation properties. The properties of the eight rainfall events studied are listed in Table 1. The rainfall events are characterized by variable rainfall depths ranging from 0.9 to 6.9 mm and mean rainfall rate from 0.4 to 11.5 mm h-1. Depending on the rainfall depths and rates, the sampling resolution was adapted. For example, R7 was collected in 22 fractions of volumes ranging from 80 to 440 mL for a rainfall depth of 3.04 mm over 30 min, while R8 was collected in 32 fractions of volumes ranging from 60 to 820 mL for a rainfall depth of 6.9 mm and lasted several hours. Note that for R7 the sampling setup allowed us to only collect the first 3.04 mm of rain of the total event (10.3 mm). Our data set consists of one (12.5 %) event with a rainfall depth of less than 1 mm, one (12.5 %) with a rainfall depth of more than 5 mm, and six others (75 %) representing rainfall depths between 1 and 5 mm. Rain events have varying cloud base heights (from 200 m for R6 up to 2000 m for R8) which, however, can fluctuate within the same event as it is the case for R8.

According to the HYSPLIT 48 h backward trajectory calculation, the origin of the air masses scavenged at the study site remained constant during the duration of the rain events, except for R6 and R8 (Fig. S3). The air masses for R1 and R2 came from the Atlantic Ocean. R3 and R4 had air masses from the Mediterranean at the surface and from Spain and Portugal at the cloud base. For the other events, influenced by mineral dust intrusion from northern Africa, the calculation of HYSPLIT backward trajectories has been performed over 120 h with the same conditions. For R5 and R6, the air masses at the surface came from the United Kingdom via the North Sea and Germany, while the air masses at the cloud base came from northern Africa (south of Tunisia or west of Libya) for R5 and from the Mediterranean Basin and Italy for R6. In the second phase of event R6 (after 09:00 UTC), the air masses at the surface also came from the Mediterranean Basin. For R7, the air masses at the cloud base came from the Mediterranean Basin, and the air masses at the surface came from Libya. For R8, the beginning of the event was characterized by air masses coming from the Atlantic through northern Morocco and Spain at the cloud base and from northern Tunisia at the surface. During the event, the origin of air masses evolved and came from different places in northern Africa (Morocco, Algeria and Tunisia), depending on altitude. This analysis of the backward trajectories shows a close alignment between the origins of these large-scale air masses and the wind directions measured at the surface at the instrumented site in Créteil.

Atmospheric aerosol mass concentrations at the beginning (average over the 30 min prior to the onset of the rainfall) of the R1, R2, R6 and R7 events are primarily controlled by PM2.5, which represents 63 %–84 % of PM10 concentrations. R3 is characterized by a lower proportion of PM2.5, which represents 38 % of PM10, while PM2.5 measured for R4, R5 and R8 correspond to 46 %–53 % of PM10. R1 to R4 took place on days with low particle concentrations, with PM10 concentrations lower than of 20 µg m-3. During these events, rain had the effect of reducing atmospheric PM10 concentrations by 11 %–53 % (Table 1). However, this illustrates the overall effect of the rain event without taking into account the increases in air concentrations that may have been observed during the events (e.g. R8). On the other hand, R5 to R7 occurred on days marked by high concentrations of both PM2.5 (33–40 µg m-3) and PM10 (47–63 µg m-3). The latter took place not only during a typical spring pollution episode (Favez et al., 2021), but also during a mineral dust intrusion from northern Africa, as shown by a multi-model dust optical depth simulation provided by the WMO Barcelona Dust Regional Centre (Fig. S4, https://dust.aemet.es, last access: 18 January 2023, Basart et al., 2019). During these events, rain was less effective at reducing PM10 concentrations. While R5 is characterized by a decrease in the PM10 concentration of the order of 17 %, R6 and R7 show no variation or an increase in the PM10 concentration (Table 1). Even though R8 occurred on a day with low particle concentrations, this event was also marked by the intrusion of mineral dust from northern Africa (Fig. S4, Table 1).

Total wet deposition fluxes in our case studies are ranging from 11 to 107 g m-2 and are not correlated with rainfall depth nor rainfall rate (Table 1). Indeed, higher wet deposition fluxes are observed for rainfall events (R5 and R6) associated with low rainfall depth but higher pre-rain PM10 concentration. However, events characterized by a similar surface PM10 mass concentration (R1, R2, R3, R4 and R8) exhibit total wet deposition fluxes that vary over a factor of 4.

The information collected makes it possible to describe eight case studies, illustrating contrasting situations in terms of meteorological conditions, dynamics and atmospheric aerosol loads.

Volume-weighted mean (VWM) mass concentrations of the particulate and dissolved phases for each rain event are represented in Fig. 2.

Regarding the particulate phase, the average mass concentrations of elements in the rainfall exhibit high variation, with values differing by a factor of 85 between events. The highest concentrations are observed for R5 and R6 events, with 33.9 and 34.5 mg L-1, respectively. Despite these fluctuations in average mass concentrations, the particulate phase is predominantly composed of Si, Fe and Al, contributing to a relative proportion between 73 % and 85 %. In contrast, the particulate Ca content displays a more pronounced variability, ranging from 3 % to 12 %, depending on the specific rain event.

Regarding the dissolved phase, R4 and R8 are the rainfall events characterized by the lowest dissolved phase VWM concentrations (∼ 2 to 3 mg L-1) and the largest rainfall amounts (4.4 and 6.9 mm for R4 and R8, respectively). These results are consistent with the dependence of wet deposition concentrations with precipitation amount and the dilution effect documented in the literature (e.g. Jaffrezo et al., 1990). The largest concentrations are of the order of 21 mg L-1 and correspond to the events marked by the mineral dust intrusion from northern Africa but also the lowest precipitation amounts (0.90 mm for R5 and 1.20 mm for R6). The rain events are not characterized by the same contents and relative proportions of acid (NO3-, SO42-) or neutralizing (NH4+) species depending on the rainfall. The dissolved phase is mainly composed of SO42-, NO3- and NH4+ (SNA), between 58 and 85 % by mass of the analysed species for R1, R2, R3, R4, R7 and R8. In contrast, R5 and R6, and to a lesser extent R8, are composed of a non-negligible proportion of Ca in the dissolved phase (23 %–40 %).

The variations in concentrations of not only acid species but also neutralizing compounds lead to different pH values in the rainfall (Table 1). The progressive scavenging of these compounds during the rainfall event also results in variations in pH (Asman et al., 1982), which is observed between the different events. For instance, R1 has a lower pH (pH < 5.6) resulting from lower average concentrations of neutralizing species. Rains R2, R3, R4, R7 and R8 have higher pH values ranging from 6.2 to 6.8 and even basic for R5 and R6 (7.5–8.0). The basic nature of R5 and R6 rains is attributed to the higher Ca contents of mineral dusts present in these rains, which is in agreement with the influence of dust intrusion, as previously described (Ma, 2006; Oduber et al., 2020).

To go further in the interpretation, EFs presented in table 2 as well as origin of air masses (Table 1), are used to classify case studies into three groups: (i) R1 to R4, characterized by air masses from the west and south of France and a significant enrichment in Ca (EF > 15), Ni (EF > 10, except R4), P (EF > 30, except R1), and very high for Zn (EF > 120) and S (EF > 1000); (ii) R5 and R6, characterized by a contribution of mineral dust and EFs reflecting mineral sources signature (between 1 and 2), except for Zn (8.0 – 13) and S (119–136), which are still lower than the other rains; and (iii) R7 and R8, characterized by low EFs (< 10) for all elements, but higher than R5 and R6 ones, except for Zn (26–44) and S (175–438).

The chemical signature allows us to classify rain events into three categories: R1, R2, R3, and R4 show a marked anthropogenic signature and are hereafter referred to as “anthropogenic” events; R5 and R6 illustrate a distinct mineral dust signature and are hereafter referred to as “mineral-dust” events; and R7 and R8 correspond to mixing conditions and are hereafter referred to as “mixed” events. However, for a given element, the EFs show that the origin is sufficiently homogeneous regardless of the rain events, limiting the data analysis as a function of aerosol sources.

The developed measurement strategy for both the chemical characterization of aerosol and wet deposition (see Sect. 2.3) enables us to compare the concentrations in the air and in the first samples of rain, excluding the effect of dilution. Total mass concentrations estimated from chemical analysis of aerosol filters represent from 15  % (R3) to 55 % (R8) of the measured PM10 mass concentration (Table 1), depending on the situations.

Total mass concentrations measured in the first fraction of rainfall events (0.06 to 0.10 mm) are higher when pre-rain PM10 surface concentrations are greater (Table 1). However, for R2, R3 and R8, PM10 concentrations are of the same order of magnitude (11.8–13 µg m-3), while total mass concentrations in their first fraction differ by a factor 1.8 (R2: 28.1 mg L-1; R3: 49.8 mg L-1; R8: 38.7 mg L-1). The latter is higher when the PM2.5 / PM10 ratio is lower (Table 1). This suggests that PM2.5 values are scavenged less effectively than coarser particles (PM2.5–10). R6 and R7 events are characterized by similar pre-rain PM10 surface concentrations as well as similar PM2.5 / PM10 ratios. However, the R6 event shows total mass concentrations in the first fraction that are 2.4 times higher than R7 (68.3 mg L-1). This can be explained by the long-range transport of mineral dust at high altitude. Therefore, wet deposition fluxes at the beginning of rainfall seem to be primarily correlated to PM10 surface concentrations and secondly to the coarse fraction (PM2.5–10 / PM10). This is consistent with the aerosol size dependence of scavenging mechanisms and the minimal efficiency of the BCS mechanism between 0.2 and 2 µm (e.g. Wang et al., 2010).

Figure 5 depicts the total concentration of elements (dissolved plus particulate) in the first rain fraction (µg kgrain-1) plotted against the total concentration of elements measured in the aerosol (µg kgair-1) for R2, R3 and R8 rain events, i.e. the only rain samplings adapted for this comparison. According to Eq. (2), the ratios in these two concentrations illustrated in Fig. 5 correspond to the WR for analysed species (Table S6 in the Supplement). It appears that the scavenging efficiency is clearly dependent on the element. As an example, for similar particulate mass concentrations (0.02 µg kgair-1), we found higher P concentration in the first fraction of R2 (88 µg kgrain-1) in comparison to Ba (14 µg kgrain-1).

We found that for a given rain event, WR values can vary by up to a factor of 11 to 30 from one element to another (Table S6). WRs of elements found in R2 are primarily in the 2000±1000 envelopes, while WRs of R3 are systematically higher. Regarding R8 events, we observed an intermediate behaviour in terms of WR values. In all the cases, the WR values are higher than the values previously estimated, in agreement with the dilution effect on the WR values available in the literature. Indeed, by taking into account the first fraction of the rainfall, the calculation minimizes the influence of the ICS contribution as opposed to the WR values considering the entire event. The difference in WR as a function of element could be due to either an additional source of elements in the rain (e.g. ICS or gas phase scavenging); a difference in BCS efficiency, e.g. due to different size distribution or hygroscopicity of the element-bearing particles; or a contribution of PM with a diameter greater than 10 µm (e.g. Jaffrezo and Colin, 1988; Cheng et al., 2021; Kasper-Giebl et al., 1999; Cheng and Zhang, 2017). Cheng et al. (2021) emphasized the predominant role of particle size distribution on the WR. Indeed, the elements associated with the coarse mode (PM2.5–10) present the largest WR, except Si and Fe, while the elements that are dominant in the fine particles (PM2.5) had a lower WR. Even if we have no information on the size distribution of aerosol chemical composition, the EF shows that the elements associated with a coarse mode by Cheng et al. (2021) are from a dust origin, and those associated with the fine mode (e.g. S, Zn) are of an anthropogenic origin, in our samples. Our results are consistent with these observations: elements linked to coarse particles, such as calcium (WR ranging between 2500 and 9800), exhibit higher WR values compared to those associated with fine particles, such as zinc (WR ranging between 1000 and 3800). However, as highlighted in the review of Cheng et al. (2021), some elements found primarily in the coarse mode, such as Fe (WR = 3800), exhibit similar WR value to elements associated with fine particles (e.g. Zn) as illustrated in event R3.

However, our study revealed a significant variation between different events for the same chemical species. Interestingly, for each element (except S), this variability consistently follows a decreasing trend in WR with increasing pre-rain PM2.5 / PM10 fraction (Table 1). In addition, we observed an increasing trend in WR with higher rainfall rates. For instance, WR of Ca increase from 2500 to 9800 when rainfall rate increases from 0.5 to 1.2 mm h-1. This shows that the particle size distribution is probably not the major factor acting on particle below-cloud wet scavenging. These results are particularly noteworthy because they represent the first instance of WR measurements unaffected by the dilution effect.

Scavenging coefficients (i.e. Λ) can be determined from the WR calculation using Eq. (3). These estimations are the first available for major and trace metals. Figure 6 illustrates Λ of elements as a function of rainfall rate. Our results show that Λ increases with rainfall rate according to a power law, as previously shown in the literature (e.g. Xu et al., 2019; Wang et al., 2014). At a rainfall rate of R= 1 mm h-1, we obtained Λ values between 0.5 and 1.3×10-6 s-1, with the exception of chlorine. These values fall within the range (2.6×10-7-1.7×10-6 s-1) documented for radionuclides by Sparmacher et al. (1993) for controlled experiments with similar rainfall rate and aerosol diameters (0.98 and 2.16 µm). Scavenging coefficient evolution with rainfall rate varies from one element to another, with slopes ranging from 0.5 for sulfur to 2.9 for chlorine. These differences cannot be attributed solely to mass concentration, particle size, or water-soluble fraction of elements. For instance, while elements associated with the same aerosol types, such as Na and Cl or Al, Ti, and Si, show similar behaviour with rainfall rate, chlorine and sulfur exhibit contrasting trends even though they are both water-soluble elements. Similarly, scavenging coefficients for coarse particles (e.g. Al and Si; 1.5–8.5×10-7 s-1) are comparable to those for fine particle (Zn and S; 0.9–6×10-7 s-1). Aerosol scavenging does not depend on a single parameter but is governed by the interaction of several parameters including the intrinsic properties of the aerosol (size,solubility) and of the precipitation (intensity, size and number of droplets). Consequently, our results underline the critical role of rainfall rates and aerosol particle properties for the determination of both WR and Λ.

The ICSC of chemical species analysed in the selected rains (see Sect. 2.4) are presented in Table 3. We observe significant variations in ICSC within individual events for different chemical species, as well as different ICSC values of the same chemical species between different events.

For R1, R2 and R4 anthropogenic events, the elements of crustal origin found in the coarse fraction of aerosols (Al, Si, Fe, Ti, Ca, Mg, Sr) are mainly deposited via the BCS mechanism. This observation is consistent with previous in situ studies conducted in urban environments, which have reported that the BCS mechanism accounts for a significant proportion (ranging from 52 % to 99 %) of calcium wet deposition (Ge et al., 2016; Karşı et al., 2018; Ge et al., 2021a; Berberler et al., 2022). The wet deposition of Mn and NH4+ is mainly attributed to the BCS mechanism, accounting for 55 % to 87 %. This corresponds to a similar range of values reported for NH4+ in other urban environments in Austria (65 %), Türkiye (60 %–95 %) and China (47 %–84 %) (Xu et al., 2017; Karşı et al., 2018; Berberler et al., 2022; Monteiro et al., 2021; Ge et al., 2021a). In the literature, the BCSC mechanism for sulfate and nitrate in urban environments shows large variations, with reported values ranging from 50 % to 98 % (Ge et al., 2016; Xu et al., 2017; Karşı et al., 2018; Ge et al., 2021a; Monteiro et al., 2021; Berberler et al., 2022) and as low as 16 % for sulfate (Aikawa et al., 2014). In our study, the BCSC of sulfate and nitrate in anthropogenic events varies between 38 % and 77 %, depending on the events. Few chemical species show a predominance of the ICS mechanism in the wet deposition of anthropogenic events that could possibly be influenced by seasonal factors, different local sources (such as oil and wood heating systems for SO42-, Zn), gas scavenging contributions (with nitrate being mainly gaseous in summer and particulates in winter) (Audoux et al., 2023), or long-distance transport. For instance, seasonal factor and difference in local sources explain higher ICSC for SO42- and Zn in R1 in comparison to R2 and R4. In addition, the higher ICSC obtained for Na and Cl for R2 may be linked to the origin of air masses coming from the Atlantic Ocean. Overall, anthropogenic events are, on average, primarily influenced by the BCS mechanism, accounting for 53 % to 77 % of the wet deposition of chemical species.

In contrast, for the mixed event R8, influenced by both local sources and long-distance transport of mineral dust, the majority of chemical species, except for V (37 %) and Cl (49 %), are predominantly deposited through the ICS mechanism, accounting for 57 % to 75 % of their wet deposition. While the long-distance transport of mineral dust may explain the pronounced contribution of the ICS mechanism for some crustal elements, it is evident that this factor alone cannot account for the prevalence of ICS for all chemical species. Certain elements observed in event R8, such as NH4+, are not associated with mineral dust. Since the rainfall depth is higher in this case, the higher ICS contribution can be due to an effective scavenging of the air column below the cloud (Ge et al., 2021a). Indeed, the wet deposition that occurs after the depletion of the atmospheric column below the cloud is primarily influenced by aerosol transported and scavenged within the cloud, explaining a high contribution of the ICS mechanism.

Several factors may contribute to these differences in the observed contribution of ICS and BCS between events. One key factor is the variation in meteorological conditions, including intensity, droplet size, and cloud base height, as well as PM10 concentrations (Table 1). Numerical studies have highlighted the importance of not only cloud height but also cloud thickness in the relative contribution of BCS and ICS (Kim et al., 2021; Migliavacca et al., 2010; Wiegand et al., 2011). This dependence can be explained by the fact that the higher the cloud height, the greater the volume of air swept by the droplets and therefore the greater the quantity of aerosols encountered by the precipitating droplets, at equal and homogeneous concentration on the atmospheric column. For example, event R1 has higher PM10 concentrations but 2 to 4 times lower rainfall depth compared to other anthropogenic events. In addition, R4 has a higher cloud base height compared to R2, which could affect the BCSC despite the higher precipitation amount and lower PM10 concentration. These variations in meteorological conditions and atmospheric dynamics could influence BCS efficiency as well as aerosol content to be scavenged below the cloud, leading to the observed discrepancies in BCSC and ICSC values. Consequently, the complex interactions between meteorological conditions, aerosol properties, local sources and long-range transport can result in different scavenging behaviours for each event, highlighting the challenge and the need for wet deposition studies.