Scavenging of Sub-Micron to Micron-Sized Microbial Aerosols during Simulated Rainfall

: The processes removing aerosols from the atmosphere during rainfall are generically referred to as scavenging. Scavenging inﬂuences aerosol distributions in the atmosphere, with consequent e ﬀ ects on cloud properties, radiative forcing, and human health. In this study, we investigated the below-cloud scavenging process, speciﬁcally focusing on the scavenging of 0.2 to 2 µ m-sized microbial aerosols by populations of water drops with average diameters of 3.0 and 3.6 mm. Rainfall was simulated in convective boundary layer air masses by dispensing the water drops from a 55 m bridge and collecting them at ground level. Particles and microbial cells scavenged by the water drops were visualized, enumerated, and sized using scanning electron and epiﬂuorescence microscopy. Aerosolized particles and DNA-containing microbial cells of 2 µ m diameter were scavenged at e ﬃ ciencies similar to those reported previously in empirical studies; however, the e ﬃ ciencies derived for smaller aerosols were signiﬁcantly higher (one to three orders of magnitude) than those predicted by microphysical modeling. Application of the derived scavenging e ﬃ ciencies to cell data from rainfall implies that, on average, approximately 50 to 70% of the 1 µ m microbial cells in the precipitation originated from within the cloud. Further study of submicron to micron-sized aerosol scavenging over a broader raindrop size distribution would improve fundamental understanding of the scavenging process and the capacity to estimate (bio)aerosol abundances in the source cloud through analysis of rainfall.


Introduction
Particles aerosolized from natural and anthropogenic sources are transported horizontally and vertically in the atmosphere, where they can have direct effects on the formation of clouds and precipitation by serving as cloud condensation and ice nuclei, respectively [1]. In-cloud water droplets that form around aerosols (i.e., in the wet phase) are deposited in precipitation together with aerosols that are scavenged by rain drops as they descend through the atmosphere. In fact, the scavenging of aerosols by rain drops is the major process of particle deposition from the troposphere [2]. As such, knowledge of wet deposition processes is vital for understanding its consequences on global distributions and concentrations of aerosols in the atmosphere.
Multiple mechanisms are involved in aerosol scavenging by rain drops, including inertial impaction, Brownian diffusion, and interception [1,3]. Aerosol scavenging simulations that consider

Simulated Rain Events
Eight simulated rain events (SREs) were generated using two commercially available~10 L watering cans (Gardener's Supply, # 06-341). The cans have interchangeable stainless-steel rose heads, which were modified so that two size populations of water drops could be generated (the holes were 1.65 ± 0.02 mm and 1.04 ± 0.02 mm in diameter). Each watering can was cleaned using detergent and rinsed thoroughly with deionized water. Prior to each experiment, the watering cans were flushed with 2 L of 0.22 µm-filtered deionized water to reduce contributions of contaminating cells and particles to the measurements. Six liters of 0.22 µm-filtered deionized water was poured from the watering cans during each experiment, and a total of~1 L of each sample was successfully recovered~2 m above ground in three 117 L galvanized cans (Home Depot, cat #001223296) that were lined with sterile polypropylene bags (Fisher #01-830E, 122 cm × 94 cm). Each SRE was approximately the same duration (mean of 41 s). After collection, the water was transferred to sterile 1-L Nalgene bottles (Nalgene #2187-0032) and chilled on ice. Immediately after return to the laboratory (~4 h), the water samples were fixed by the addition of formalin to a final concentration of 4% (v/v), stored at 4 • C, and processed within 30 days of collection.
To determine the mean volumes of the drop populations, two of the SREs (7 and 8, watering cans with 1.65 ± 0.02 mm and 1.04 ± 0.02 mm diameter holes, respectively) were collected in a dewar containing liquid nitrogen [23] and processed as described previously [20]. Briefly, the frozen water drops collected in the liquid nitrogen were recovered and placed into individual microcentrifuge tubes. These drops were then thawed and their volume was determined using a micropipette [23].
Procedural blanks were prepared for each SRE and analyzed in parallel to assess the level of background microbe and particle contamination associated with the measurements. The controls used the same water source as the SREs but were poured directly into a bag-lined galvanized can, collected, and analyzed.

Measurement of Ambient Particulate Matter
Two Plantower PMS7003 particle sensors and two Nova Fitness SD021 particle sensors were positioned adjacent to where the SREs were deployed from the top of the bridge. Both sensors measure particulate matter between 0.3 and 10 µm in diameter. Data were continuously collected during the experiments, with the Plantower PMS7003 measuring ambient particulate matter (PM) ≤1, ≤2.5, and ≤10 µm in diameter, and the Nova Fitness SD021 measuring PM ≤ 2.5, and ≤10 µm. The particulate matter data are expressed by weight (µg m −3 ).

Characterization of Microbial Cells and Particles
To enumerate DNA-containing cells, triplicate 10 mL samples from each SRE and control were stained with a final concentration of 25× SYBR TM Gold (Life Technologies Corp., cat. no. S-11494) in the dark for 15 min. The samples were then filtered onto black, 25 mm diameter 0.22 µm polycarbonate Isopore™ filters (Millipore, cat. no. GTTP04700). A 4 µL drop of antifade (1:1 solution of glycerol and PBS, 0.1% phenylenediamine) was added to each of the filters before coverslips were mounted. A Nikon ECLIPSE Ni epifluorescence microscope was used to visualize and enumerate the DNA-containing microbial cells in sixty random fields of view (FOV, area of 1.8 × 10 4 µm 2 ) on each filter. NIS-Elements Advanced Research software (Nikon, Inc., Tokyo, Japan) was used to automatically trace and measure the cells in each FOV. Cells that were present as aggregates were not individually counted; each aggregate was sized and considered a single particle [24].
For particle characterization, triplicate 10 mL samples from eight SREs and controls were filtered onto white, 25 mm diameter 0.22 µm polycarbonate filters (GE Water & Process Technologies, cat. no. K02CP02500) and sputter coated with gold palladium for 1.5 min [12]. Particles were visualized using a Hitachi SU5000 Schottky Field-Emission scanning electron microscope under high vacuum mode. Sixty random FOVs (69 µm 2 each) from each filter were imaged at a voltage of 11-12 kV and the particle area was traced and analyzed using ImageJ software [12,25]. Particles that appeared aggregated were counted as a single particle [24].
Particle and cell area was used to calculate equivalent diameters according to the following formula [24]: Particle and cell concentration per SRE sample was calculated based on the average number of objects per FOV, filtration area (226.9 mm 2 ), and sample volume (10 mL). Particles and cells with Atmosphere 2020, 11, 80 4 of 14 equivalent diameters of 0.2 to 2 µm were binned into 10 categories of 0.2 µm increments. For each size bin where the sample and control concentrations were significantly different, the mean number of objects (particles or cells) per FOV in the controls were subtracted from the mean number of objects per FOV in the samples. The binned number concentrations were used to estimate particle and cell mass by assuming a particle density of 1.0 g cm −3 and cell density of 1.1 g cm −3 [26,27].

Ambient Particulate Size Distribution
The size distribution of ambient aerosols 0.2 to 2 µm in diameter was estimated by Equation (2) [28]: where n is the number of particles cm −3 of radius r (µm) and Φ is the volume of particles per unit volume of air sampled. This equation is based on Junge's power law [29] and creates size distribution curves that have been shown to agree well with empirical measurements of aerosols with radii >0.05 µm [24,28,29]. PM 10 measurements taken during each SRE (W) and the particle density estimated per unit volume (ρ) were utilized in the calculation of Φ (Φ = W/ρ; [28]). The mass concentration distribution of ambient particles was inferred from the number concentration distribution and assuming a density of 1.0 g cm −3 [27]. Cell mass and number distribution was estimated from the ambient particle distribution, the cell to particle ratio observed by bin in the 3.6 mm drops (Tables S1 and S2), and assuming a cell density of 1.1 g cm −3 [26].

Application of the Scavenging Efficiencies to Rain Data
The capture cross-section of a droplet, namely E × π × D 2 /4 [21], was used to estimate the volume of air scavenged by a drop (V) falling from a cloud-base height of B (in meters). This was calculated according to Equation (4): The number of particles or cells removed from the air column below the cloud by a single drop was estimated by multiplication of V by the number of ambient particles or cells m −3 .
To assess the contribution of scavenging to the microbes deposited in rain at the surface, the fraction of cells scavenged and that originated from the precipitating cloud was estimated using select events from a two-year record of rainfall data from Baton Rouge, Louisiana [30]. Following Marshall and Palmer's [31] drop-size distributions derived from rain rates, 13 rain events were chosen for the scavenging calculations, representing end members of rain rates (mean 2.8 mm h −1 , n = 7; and 14.5 mm h −1 , n = 6) that would be expected to contain drop sizes relevant to this study (average equivalent diameter of 3.0 and 3.6 mm). The cloud-base heights for both subsets of storms ranged from 100 to 1160 m above ground level, and the cell concentrations in the rain samples ranged from 8.45 × 10 4 to 1.3 × 10 6 cells L −1 [30]. Using Equation (2), the number of 1 µm bioaerosols (in cells m −3 ) for each rain event was estimated based on the 24-h average PM 10 records for Baton Rouge (US Environmental Protection Agency daily records, https://www.epa.gov/outdoor-air-quality-data). The estimated number of cells scavenged by each rain drop size was based on the cloud-base height for each rain event, the capture cross-section of each raindrop, and the concentration of 1 µm bioaerosols Atmosphere 2020, 11, 80 5 of 14 (Equation (4)). The difference between the number of scavenged cells per liter of drops and the cell density in each rain sample provides an estimate for the number of cells that originated from the cloud (As compared to below-cloud).

Statistical Analyses
R Software Version 3.5.1 [32] was used for all statistical analyses. Differences between concentration, volume, and efficiencies were determined using Student's t-tests (base R). Analysis of variance (ANOVA) was utilized to test the differences between measured and calculated ambient particle concentrations. Scavenging efficiency models were produced and plotted with the ggplot2 package [33]. The models were fit with the stat_smooth function using the linear model method and the y = x + x 2 formula [34].

Water Drop Volumes
Hundreds of drops generated from two dispensing cans with different hole sizes (1.04-and 1.65-mm diameter) were recovered in liquid nitrogen from SRE 7 and 8 ( Table 1). From these populations, the volume of randomly selected drops from SRE 7 (n = 56) and SRE 8 (n = 52) was measured. Mean drop volumes of 14.6 ± 1.6 µL and 25.0 ± 1.9 µL (± standard error of the mean) from the small and large holes, respectively, in the dispensing cans were significantly different (t-test, p < 0.001). Assuming a sphere, the drop volumes correspond to equivalent diameters of 3.0 ± 1.5 mm and 3.6 ± 1.5 mm ( Figure 1).
Atmosphere 2020, 11, x FOR PEER REVIEW 5 of 13 variance (ANOVA) was utilized to test the differences between measured and calculated ambient particle concentrations. Scavenging efficiency models were produced and plotted with the ggplot2 package [33]. The models were fit with the stat_smooth function using the linear model method and the y = x + x 2 formula [34].

Water Drop Volumes
Hundreds of drops generated from two dispensing cans with different hole sizes (1.04-and 1.65mm diameter) were recovered in liquid nitrogen from SRE 7 and 8 ( Table 1). From these populations, the volume of randomly selected drops from SRE 7 (n = 56) and SRE 8 (n = 52) was measured. Mean drop volumes of 14.6 ± 1.6 µL and 25.0 ± 1.9 µL (± standard error of the mean) from the small and large holes, respectively, in the dispensing cans were significantly different (t-test, p < 0.001). Assuming a sphere, the drop volumes correspond to equivalent diameters of 3.0 ± 1.5 mm and 3.6 ± 1.5 mm ( Figure 1).

Ambient Particulate Size Distribution
Ambient PM 10 concentrations measured during the experiments were not statistically different (t-test, p > 0.05) and averaged 22 ± 4 µg m −3 . Using this value and Equation (2), the number distribution of particles m −3 was determined ( Figure 2). The concentration of ambient particles and cells were estimated from this distribution as described in Section 2.5 ( Figure 2).

Ambient Particulate Size Distribution
Ambient PM10 concentrations measured during the experiments were not statistically different (t-test, p > 0.05) and averaged 22 ± 4 µg m −3 . Using this value and Equation (2), the number distribution of particles m −3 was determined ( Figure 2). The concentration of ambient particles and cells were estimated from this distribution as described in Section 2.5 ( Figure 2). The PM10 mass concentration distribution (Figure 2) in the sampled air masses are consistent with the Environmental Protection Agency Air Quality Index category "Good" and indicative of relatively low anthropogenic emissions. Inferred biomass was highly similar across the size bins, but inferred cell concentrations were highest for cells <1 µ m in diameter. The estimated µ g m −3 distribution (Figure 2) was not statistically different (ANOVA, p = 0.99) from measured values of PM10, PM2.5, and PM1 sampled during the SREs ( Figure S1). Based on these results, the calculated µ g m −3 distribution was used in the scavenging efficiency calculations.

Cell and Particle Scavenging
Based on data from the procedural controls and a limit of detection at 3-sigma, our method allowed detection of 9.0 × 10 2 cells and 6.5 × 10 5 particles mL −1 . The concentrations of particles (raw mean 1.0 ± 0.18 × 10 6 mL −1 ) and cells (raw mean 2.0 ± 0.049 × 10 4 mL −1 ) in samples for the 3.6 mm drops were above the level of detection and significantly different (t-test, p < 0.001) from the procedural controls (background average 2.1 ± 1.1 × 10 2 cells and 3.5 ± 0.70 × 10 5 particles mL −1 ). For samples of the 3.0 mm drops, the raw concentrations of cells (raw mean 1.3 ± 0.18 × 10 3 cells mL −1 ) were also The PM 10 mass concentration distribution (Figure 2) in the sampled air masses are consistent with the Environmental Protection Agency Air Quality Index category "Good" and indicative of relatively low anthropogenic emissions. Inferred biomass was highly similar across the size bins, but inferred cell concentrations were highest for cells <1 µm in diameter. The estimated µg m −3 distribution (Figure 2) was not statistically different (ANOVA, p = 0.99) from measured values of PM 10 , PM 2.5 , and PM 1 sampled during the SREs ( Figure S1). Based on these results, the calculated µg m −3 distribution was used in the scavenging efficiency calculations.

Cell and Particle Scavenging
Based on data from the procedural controls and a limit of detection at 3-sigma, our method allowed detection of 9.0 × 10 2 cells and 6.5 × 10 5 particles mL −1 . The concentrations of particles (raw mean 1.0 ± 0.18 × 10 6 mL −1 ) and cells (raw mean 2.0 ± 0.049 × 10 4 mL −1 ) in samples for the 3.6 mm drops were above the level of detection and significantly different (t-test, p < 0.001) from the procedural controls (background average 2.1 ± 1.1 × 10 2 cells and 3.5 ± 0.70 × 10 5 particles mL −1 ). For samples of the 3.0 mm drops, the raw concentrations of cells (raw mean 1.3 ± 0.18 × 10 3 cells mL −1 ) were also above the level of detection and significantly different from the background concentrations; however, the particle data (raw mean 5.8 ± 1.1 × 10 5 particles mL −1 ) were only above the level of detection in the 0.6 to 0.8 and 1.4 to 1.6 µm diameter size bins. Background concentrations were subtracted from the raw sample concentrations for all size bins that were above the level of detection (Figure 3). above the level of detection and significantly different from the background concentrations; however, the particle data (raw mean 5.8 ± 1.1 × 10 5 particles mL −1 ) were only above the level of detection in the 0.6 to 0.8 and 1.4 to 1.6 µ m diameter size bins. Background concentrations were subtracted from the raw sample concentrations for all size bins that were above the level of detection (Figure 3).  Tables S1and S2). This equates to 1.0 ± 0.29 × 10 4 and 3.0 ± 0.55 × 10 4 particles scavenged per 3.0-or 3.6-mm drop (Table  1), respectively, while descending 53 m in the atmosphere. Bulk samples of the 3.0-and 3.6-mm drops contained 0.4 to 2 µ m cells at average concentrations of 9.9 ± 2.7 × 10 2 and 1.8 ± 0.42 × 10 3 cells mL −1 , respectively ( Figure 3, Tables S1and S2). This corresponds to 68 ± 18 and 73 ± 17 cells scavenged per 3.0-or 3.6-mm drop, respectively, during the experiments.
The scavenging efficiencies of particles and cells collected by the 3.6 mm drops showed an identical trend, with minimum values observed for the smallest particle and cell diameters (Figure 4, Tables S1and S2). The scavenging efficiencies of 3.6 mm drops for cells and particles with mean diameters 0.5 to 1.9 µm were not significantly different (p > 0.05) (Tables S1and S2). Similarly, scavenging efficiencies of 3.0 mm drops were not significantly different between cells and particles within the 0.6 to 0.8 and 1.4 to 1.6 µ m bins (p > 0.05) (Tables S1and S2). Therefore, the particle and cell data were combined and fitted to quadratic models to examine the difference in scavenging efficiencies between the 3.0-and 3.6-mm drops (Figure 4). The highest scavenging efficiencies for both drop sizes were for the largest (2 µ m) particles, and the efficiency decreased exponentially with particle size. The efficiency at which 3.6 mm drops scavenged particles in the 0.4 to 2 µ m size range was ~twofold higher than the 3.0 mm drops. Particle abundance between individual size bins was significantly different based on drop size (t-test, p < 0.01). The concentrations of 0.2 to 2 µm particles collected by the 3.0-and 3.6-mm drops averaged 1.4 ± 0.42 × 10 5 and 7.3 ± 1.4 × 10 5 particles mL −1 , respectively (Figure 3, Tables S1 and S2). This equates to 1.0 ± 0.29 × 10 4 and 3.0 ± 0.55 × 10 4 particles scavenged per 3.0-or 3.6-mm drop (Table 1), respectively, while descending 53 m in the atmosphere. Bulk samples of the 3.0-and 3.6-mm drops contained 0.4 to 2 µm cells at average concentrations of 9.9 ± 2.7 × 10 2 and 1.8 ± 0.42 × 10 3 cells mL −1 , respectively (Figure 3, Tables S1 and S2). This corresponds to 68 ± 18 and 73 ± 17 cells scavenged per 3.0-or 3.6-mm drop, respectively, during the experiments.
The scavenging efficiencies of particles and cells collected by the 3.6 mm drops showed an identical trend, with minimum values observed for the smallest particle and cell diameters (Figure 4, Tables S1 and S2). The scavenging efficiencies of 3.6 mm drops for cells and particles with mean diameters 0.5 to 1.9 µm were not significantly different (p > 0.05) (Tables S1 and S2). Similarly, scavenging efficiencies of 3.0 mm drops were not significantly different between cells and particles within the 0.6 to 0.8 and 1.4 to 1.6 µm bins (p > 0.05) (Tables S1 and S2). Therefore, the particle and cell data were combined and fitted to quadratic models to examine the difference in scavenging efficiencies between the 3.0-and 3.6-mm drops (Figure 4). The highest scavenging efficiencies for both drop sizes were for the largest (2 µm) particles, and the efficiency decreased exponentially with particle size. The efficiency at which 3.6 mm drops scavenged particles in the 0.4 to 2 µm size range was~twofold higher than the 3.0 mm drops. Atmosphere 2020, 11, x FOR PEER REVIEW 8 of 13  To explore the contribution of scavenged bioaerosols to cell concentrations in precipitation, we used select events from a published precipitation dataset collected in Louisiana [30]. We assumed the majority of bacterial-sized aerosols were 1 µ m in diameter and used the corresponding modelled scavenging efficiencies for 3.0-and 3.6-mm drops to infer source contributions from scavenging versus the precipitating cloud (Table 2). Low intensity rain events are associated with smaller drops (1 to 2 mm), whereas moderate to heavy rainfalls can produce larger drops (1 to 4 mm) [2]. Our analysis implies higher scavenging of cell sizes within the scavenging gap range for larger drop sizes and rainfall intensities. For example, calculations based on the lowest intensity rainfalls (2.8 mm h −1 ) suggest 27% of cells in precipitation were scavenged by drops below the cloud, whereas higher intensity rainfalls (14.5 mm h −1 ) had higher rates of scavenging (45%; Table 2). Nevertheless, our data and related calculations imply that the majority of cells in these rain samples originated from the source cloud. Table 2. Effect of drop size and rain intensity on cell scavenging. The proportion of cells in Louisiana rain samples that were scavenged below-cloud (%S) or originated from the cloud (%C) were estimated using the scavenging efficiencies in Figure 4. Rain data are from [29].

Discussion
A number of phoretic forces are known to affect scavenging of particles within the scavenging gap range. The negative surface charge possessed by bacterial cells could potentially influence drop scavenging through electrophoresis [6]. Nonetheless, we observed no significant differences in drop scavenging of microbial cells and particles with mean equivalent diameters of 0.5 to 1.9 µ m ( Figure  4). Diffusiophoresis can occur during drop evaporation, with a higher concentration of water vapor nearest the evaporating drop resulting in particle motion away from the drop [1]. Thermophoresis is also possible if there is a temperature gradient between the drop and the surrounding air. For To explore the contribution of scavenged bioaerosols to cell concentrations in precipitation, we used select events from a published precipitation dataset collected in Louisiana [30]. We assumed the majority of bacterial-sized aerosols were 1 µm in diameter and used the corresponding modelled scavenging efficiencies for 3.0-and 3.6-mm drops to infer source contributions from scavenging versus the precipitating cloud (Table 2). Low intensity rain events are associated with smaller drops (1 to 2 mm), whereas moderate to heavy rainfalls can produce larger drops (1 to 4 mm) [2]. Our analysis implies higher scavenging of cell sizes within the scavenging gap range for larger drop sizes and rainfall intensities. For example, calculations based on the lowest intensity rainfalls (2.8 mm h −1 ) suggest 27% of cells in precipitation were scavenged by drops below the cloud, whereas higher intensity rainfalls (14.5 mm h −1 ) had higher rates of scavenging (45%; Table 2). Nevertheless, our data and related calculations imply that the majority of cells in these rain samples originated from the source cloud. Table 2. Effect of drop size and rain intensity on cell scavenging. The proportion of cells in Louisiana rain samples that were scavenged below-cloud (%S) or originated from the cloud (%C) were estimated using the scavenging efficiencies in Figure 4. Rain data are from [29].

Discussion
A number of phoretic forces are known to affect scavenging of particles within the scavenging gap range. The negative surface charge possessed by bacterial cells could potentially influence drop scavenging through electrophoresis [6]. Nonetheless, we observed no significant differences in drop scavenging of microbial cells and particles with mean equivalent diameters of 0.5 to 1.9 µm (Figure 4). Diffusiophoresis can occur during drop evaporation, with a higher concentration of water vapor nearest the evaporating drop resulting in particle motion away from the drop [1]. Thermophoresis is also possible if there is a temperature gradient between the drop and the surrounding air. For example, aerosols may be attracted toward an evaporating drop that becomes colder at its surface. However, phoretic effects are generally thought to be negligible unless there is a steep temperature or gas gradient (e.g., >1 • C difference between the drop surface and air temperature) [1,35]. Brownian diffusion may also effect scavenging by bringing a particle into contact with a drop, and this random motion is important for the scavenging of particles <0.2 µm in diameter [1]. For aerosol particles >2 µm in diameter, chemical composition may affect scavenging rates; however, its effects may be negligible for the particle sizes examined in this study [36]. Considering this, the most relevant scavenging mechanism applicable to particles with equivalent diameters of 0.2 to 2 µm is interception.
Drop scavenging of particles in the scavenging gap size range ( Figure 4) occurs at much higher efficiencies than the model predictions of Quérel et al. [22] and Slinn [3] ( Figure S2). The scavenging efficiencies derived for the 3.0 and 3.6 mm drops were at least an order of magnitude greater than those predicted in the DESCAM model ( Figure S2; DEtailed SCavening Model, [11,22]). For example, the DESCAM model efficiency for 0.81 µm diameter particles is 9.30 × 10 −4 for 3.25 mm drops, whereas observed efficiencies for 3.0 and 3.6 mm drops are 2 × 10 −2 and 4 × 10 −2 , respectively ( Figure S2) [22]. The empirically derived drop scavenging efficiencies are also~600-fold higher than those for 3.6 mm drops using the interception term from the Slinn model ( Figure S2) [37]. This is consistent with previous studies that have observed scavenging efficiencies for particles within the scavenging gap one to five orders of magnitude greater than efficiencies predicted using the Slinn model [7,10,37,38]. Volken and Schumann [39] and Chate and Pranesha [10] suggest that some of the mechanisms neglected in current models (e.g., phoretic forces and the growth of hygroscopic aerosol particles) may explain the discrepancies. Nevertheless, attempts to incorporate parameters that may be relevant (i.e., thermophoresis, diffusiophoresis, and electric forces) have not improved the predictive value of models for particles within the scavenging gap [10].
Rain drops in many precipitation events are 1 to 3 mm in diameter, with larger drops typically forming from melting ice particles produced by convective storms [2,31]. The drop diameters (3.0 and 3.6 mm) examined in this study are relevant for low to moderately intense rainfall rates in the range of 2.5-25 mm h −1 [31,40]. There have been few empirical data available on scavenging within the scavenging gap by drops of known diameters. For example, our results generally agree with those measured during natural rainfall events [6,7]; however, the lack of drop diameter data makes direct comparisons difficult. Also notable is the work of Pranesha and Kamra [41], who examined large water drops (3.6, 4.2, and 4.8 mm diameter) and the scavenging of particles 1.9 to 6.4 µm diameter. The scavenging efficiency they found for 3.6 mm drops of 1.9 µm diameter particles (~3 × 10 −1 ) is similar to our estimate (2 × 10 −1 ; Figure S2). Although below-cloud scavenging of bacterial and fungal bioaerosols has been studied previously [20], there have been no measurements of drop scavenging efficiencies for particles that are representative of bacterial cell sizes <2 µm.
Based on the drop population sizes tested in this study, scavenging efficiency increases with drop diameter (Figure 4). This relationship has been observed previously [42][43][44], but it is not consistent with all studies. For example, Pranesha and Kamra [41] observed a decrease in collection efficiency of particle diameters 1.9 to 6.4 µm with increasing drop size (3.6-4.8 mm). The Slinn model also predicts a decrease in collection efficiency with increasing drop size [37]. However, Hampl et al. [44] found that collection efficiency increased with increasing drop diameter (1.44 to 5.08 mm) for silver chloride aerosols with average diameters of 0.366 µm. The experimental data of Beard et al. [42] for scavenging of 0.38 to 4.4 µm particles by 0.28 to 1.24 mm drops also demonstrated increased efficiency with drop size. Furthermore, there were no discernable differences in the experimental collection efficiencies measured by Quérel et al. [22] for 2.0 to 2.62 mm drops, suggesting little to no difference between drop sizes. This collection of contrary results, generated under a variety of experimental conditions (e.g., temperature, relative humidity, drop charge, aerosol composition and diameter), makes clear the complexity of this phenomenon and need for future studies that examine scavenging under controlled experimental conditions [4,38].
In addition to improving understanding of bioaerosol removal by descending rain drops in the atmosphere, information about the scavenging process can be used to assess the fraction of particles in rain that originated from the source cloud. We used data on microbial cell concentrations from Louisiana rain [30] and the scavenging behaviors documented in this study to infer that the majority (55% to 73%) of 1 µm microbes in low to moderate rainfall intensity events originated from within the cloud (Table 2). When compared to low intensity rainfall, moderately intense rainfall produces larger drop sizes, which tend to be more efficient at scavenging below-cloud aerosols (Figure 4). Our results are consistent with recent observations showing the fungal composition of rain correlated with lower altitude characteristics of storms, whereas the smaller-sized bacteria assemblages correlated with macroscale drivers that implied nonlocal sources [45]. It should be noted our calculations have ignored the contribution of microbial aerosols within the cloud that are subject to scavenging as drops descend through the cloud. As such, we are likely overestimating the fraction of aerosols that are sourced from the wet phase of a cloud. Furthermore, since we did not consider the progressive reduction of aerosol concentration during each storm, the total aerosols scavenged per rain event are likely overestimated.
Various studies have leveraged the scavenging gap in attempts to estimate the concentration of particles and microbial cells in cloud water based on precipitation data [19,30,46]. If, however, the contribution of below-cloud scavenging is much larger than currently appreciated, then the cloud contribution of particles and cells will be overestimated [19,46] and assumptions inherent to aerosol transport models [14,47] may be flawed. For example, the particle transport model utilized by Burrows et al. [48] incorporated Slinn model scavenging efficiencies on microbial-sized particles [49], which are roughly three orders of magnitude lower than the values we obtained ( Figure S2). Hence, improved understanding of droplet scavenging efficiency for particles with diameters of 0.2 to 2 µm would allow atmospheric residence time estimates for microbial aerosols to be constrained with greater confidence.

Conclusions
Although we did not determine the composition of microorganisms or particles present in the air masses sampled, our results ( Figure 4) indicate that microbial aerosol composition was less important than their equivalent diameter. Therefore, we speculate that altering the microbial species that compose the bioaerosols would not have a perceptible effect on scavenging. These experimental results are consistent with the limited number of data available from empirical studies (e.g., [41]), documenting order of magnitude discrepancies between experimentally-derived scavenging efficiencies and those widely used in simulation models [22,37]. Intrinsically, the assumptions on which a number of microphysical models are based severely underestimate PM 2.5 deposition flux for wet processes. New observations of rain scavenging over a broader range of drop diameters would improve understanding of how precipitation that deposits microorganisms entrained in cloud water droplets affects aerosol distributions in the air masses underlying precipitating clouds, and controls removal of atmospheric particles that have important consequences on public health. To our knowledge, no studies have considered scavenging processes for submicron aerosols over the entire range of drop diameters found in rain (0.5-5 mm) [2]. Future investigations that examine the effect of environmental conditions (e.g., temperature and relative humidity), aerosol composition, and a wider range of drop sizes could lend valuable insight into the wet deposition process for aerosol sizes within the scavenging gap.