Field experiments were conducted at a saltwater pond, the Great Pond, Falmouth, MA, United States (41.5580N 70.5841W) (Figure 1, top left), and a freshwater lake, Claytor Lake, Dublin, VA, United States (37.0530N 80.6208W) (Figure 1, top right and bottom). The Great Pond is a large salt water pond that is connected to the ocean. Claytor Lake is a freshwater reservoir fed by the New River.

An electric Kingfisher USV (Clearpath Robotics, Kitchener, ON, Canada) served as the base sampling platform for collecting and measuring bioaerosols in both aquatic environments. This Kingfisher USV has a catamaran hull and was approximately 30 kg in weight with payload, with dimensions 1.35 m × 0.98 m × 1.2 m (L × W × H), as equipped for use in this study. The Kingfisher was operated via remote control. Batteries provided about 3 h of runtime and could be exchanged when needed. The propulsion system was a differential thrust system driven by two impellers inside each of the two hulls giving a maximum speed of 1.7 m/s, allowing for precise movements in shallow waters with little disturbance to the surrounding aquatic environment that could affect sampling. Six 30-min USV sampling missions were conducted at Great Pond, and six 30-min USV sampling missions were conducted at Claytor Lake. For the Great Pond experiments, sampling missions were conducted from about 10:00 AM to 15:00 PM EST on August 20, 2017 (Table 1). For the Claytor Lake experiments, sampling missions occurred from about 11:30 AM to 15:30 PM EST on October 3, 2017 (Table 1). Each sampling mission consisted of loading six impingers with autoclaved and 0.2 μm filtered media, alternating R2 broth, or R2B (R2 Broth Media 3.15 g/L, Teknova Cat. No. R0005, Hollister, CA, United States), and MO broth, or MOB [Modified Ocean Broth Media 18 g/L Instant Ocean (Spectrum Brands, Blacksburg, VA, United States), 1 μM final NH₄Cl, 0.1 μM final KH₂PO₄] (Reasoner and Geldreich, 1985; Rappé et al., 2002; Joint et al., 2010). The USV then transitioned to the sampling location using visual markers such as buoys for guidance. Once in place, the USV held position using thrusters to stay pointed into the wind and on location. The sampling platform was extended to the vertical position. Data collection from particle and meteorological sensors was started. Vacuum pumps for the appropriate impingers were turned on. Sampling at a rate of 1.6–2.4 LPM was continued for 30 min. At the conclusion of the 30-min sampling duration, the vacuum pumps were turned off and recording of particle and meteorological data was completed. The sampler platform was returned to the stowed configuration and the USV transitioned back to the shore. Impingers were removed, and the 50 mL conical tubes were unscrewed from the impinger body and capped with sterile lids. The impinged samples were then placed on ice and transported to the laboratory for processing after the conclusion of all sampling missions for the day.

A small unmanned aircraft system (sUAS) was used to collect microorganisms in the lower atmosphere as described by Jimenez-Sanchez et al. (2018). Flights were coordinated within the USV sampling periods such that both vehicles were sampling simultaneously. The sUAS flew an orbital (circular) pattern with a target altitude of 50 m above ground level (AGL), with the USV at the approximate center of the orbit. The sampling device was closed during takeoff and landing, and was opened by remote control from the ground once the sUAS was at the target sampling altitude and airspeed. The device remained open for the duration of the 10-min sampling interval. Immediately following sample collection, the exposed plate containing agar media [alternating R2A (R2 as described above with 15% agar, Thermo Fisher Scientific Cat. No. 9002-18-0, Asheville, NC, United States) and MOA (MO as described above with 5.8 g/L Gelzan gelling agent, PhytoTechnology Laboratories Cat. No. G3251, Lenexa KS 66215)] was removed from the sampling device and stored in a small plastic container for transport to the laboratory. Unexposed (control) plates were placed in the same plastic container, and incubated under the same conditions as sUAS samples. Flights were conducted by a Federal Aviation Administration (FAA)-certified pilot (Schmale) under Remote Pilot Certificate Number 4038906 with an observer (McClelland). Flights over Claytor Lake, Dublin, VA, United States were conducted under Virginia State Parks Special Use Permit Number 4-012-017 issued by Chris Doss, Park Manager.

A bioaerosol-sampling system was integrated into the USV to sample at different heights above the water. The system was designed to extend the water-sampling capabilities of this platform (Powers et al., 2018), though we did not collect water samples as part of the field campaigns described in this manuscript. Four optical particle counters from two different manufactures, the SDS021 (Nova Fitness Co., Ltd., Jinan, Shandong Province, China) and the PMS7003 (Plantower, Shunyi District, Beijing, China), were used (Figure 2). One of each type of sensor was placed at 0.1 and 1.1 m above the water, so that at each height, there were two different types of sensors (Figure 2). The SDS021 reported PM2.5 and PM10 concentrations while the PMS7003 reported PM1, PM2.5, and PM10 concentrations. These sensors collected measurements at a sampling rate of about 1 Hz. In addition to the PM counters, a custom impinger was designed and printed from high-density polyethylene (which allowed the impinger to be autoclaved) (see Supplementary Files for 3D-printing (.stl) files here: https://github.com/SchmaleLab/Schmale-Lab-3D-Printing-Files-Powers-et-al-Frontiers-2018) (Figure 3). The impinger was designed to be screwed directly onto a 50 mL sterile polypropylene conical centrifuge tube that served as the collection vessel (Thermo Fisher Scientific Cat. No. 05-538-60, Asheville, NC, United States). The impinger was also designed around disposable borosilicate Pasteur pipettes (Corning CLS7095D5X, SIGM-ALDRICH, St. Louis, MO, United States) serving as the down tube that air would travel through into the impinger liquid (Figure 3). A 1 cm diameter copper plumbing elbow served as the air inlet to the impinger (Figure 3). Lin et al. (1997) tested different impinger designs, which guided the design of the impinger used in this study to ensure the highest possible collection efficiency. Three of these impingers where used in two groups at each height that included two PM counters (one of each manufacturer) with one sensor group at 0.1 m above the water surface [Figure 2C, impingers 4 (i4), 5 (i5) and 6 (i6) from left to right] and the second sensor group at 1.1 m above the water surface [Figure 2B, impingers 1 (i1), 2 (i2) and 3 (i3) from left to right]. Each impinger was connected to a pump modified to operate as a vacuum pump (ZT370-01, Dongguan Zhentian Precision Electronics Co., Ltd., Dongguan city, Guangdong province, China) A flowmeter was used to determine the sampling rate of each vacuum pump (L/min across three independent testing cycles). Impingers i1 and i6 had a mean rate of 1.0 and 1.03 L/min, respectively; a combined sampling rate of 2.03 L/min was used after i1 and i6 sample volumes were combined (i1 and i6 each contained 20 mL of MOB media). Impingers i3 and i4 had a mean rate of 0.61 and 1.27 L/min, respectively; a combined sampling rate of 1.88 L/min was used after i3 and i4 sample volumes were combined (i3 and i4 each contained 20 mL of R2B media). Air was pulled into the impinger nozzle down the pipette tube where it then traveled through 20 mL of liquid media in the conical tube. Control tubes i2 and i5 contained 20 mL of MOB and R2B, respectively, for even numbered missions and 20 mL of R2B and MOB, respectively, for odd numbered missions. These sensor groups were mounted on a two vertical carbon fiber tubes attached to a carbon fiber tubing base attached to the USV (Figure 2). The carbon fiber base allowed the vertical sensor assembly to be extended forward in an arc about 45 degrees aft via a stepper motor screw assembly for safe transport between sampling missions. The sensors, vacuum pumps, and stepper motor where connected to a microcontroller (Teensy 3.6, PJRC.COM, LLC., Sherwood, OR, United States) that controlled all sensor operations. An Airmar 200WX marine weather station was integrated into the USV to capture environmental data including wind speed, wind direction, and temperature. All sensor data was transmitted to the USV computer via serial communications for recording. Sensor actuation and data collection were controlled through the USV computer by a command computer on shore over a 2.4 GHz WiFi data link.

Agar collection plates from the sUAS sampling missions contained R2A or MOA. R2A medium was used to favor the growth of freshwater bacteria, and MOA medium was used to favor the growth of saltwater bacteria. R2A plates were incubated for 3–4 days at room temperature, MOA plates were incubated for 4–10 days at 15°C, and colony forming units (CFUs) were counted and the plates were photographed. A subset of the colonies on each plate were picked with sterile toothpicks and inoculated into 140 μL of H₂O for ice nucleation assays and subsequent storage at minus 80°C in 20% glycerol. CFU per plate per sUAS flight were converted to CFU/m³ of air sampled as per (Aylor et al., 2011; Lin et al., 2014).

Polystyrene latex (PSL) beads with mean sizes of 3.0 μm (cat #LB30-1ML, Sigma) and 1.0 μm (cat #89904, Sigma), similar in size to bacteria, were chosen to determine impinger collection efficiency and the accuracy of the optical particle counters relative to the APS. A nebulizer, supplied with HEPA filtered air through a hydrocarbon trap, was used to aerosolize the PSL beads. Following aerosolization, the aerosols were routed through a dryer (TSI Model 3062, Shoreview, MN, United States), a Kr-85 neutralizer (TSI, Shoreview, MN, United States), and into a 280L chamber (AtmosBag two-hand, size M, Cat. #Z530212, Sigma-Aldrich, St. Louis, MO, United States). An impinger (described above) containing 30 mL of filter sterilized water (Millipore # GSWP04700), two optical particle counters (SDS021 and PMS7003), and a fan (to promote mixing) were placed inside the chamber. PSL beads were nebulized for 2 min to fill the chamber, and the chamber was allowed to equilibrate for 5 min before sampling. The aerosol size distribution and concentration were determined with an Aerodynamic Particle Sizer (TSI Model 332100, Ser. #71102298, Shoreview, MN, United States) during 2-min sampling runs. This experiment was conducted in triplicate. Collection efficiency (%) was calculated as the [(C_atmosbag - C_{impingeroutlet})/C_atmosbag]×100, where C is the particle concentration of the PSL beads.

Statistical analyses were conducted using R version 3.4.2. A linear regression model was used to examine differences among culturable bacteria collected during 12 sampling missions over 2 days with meteorological and PM concentration data. A circular linear regression model was used to compare wind direction with PM concentration data. A 95% confidence interval was used for significant differences (P < 0.05).

Six USV missions and four sUAS missions were conducted on August 20, 2017 at the Great Pond, Falmouth, MA, United States (Tables 1, 2). The concentration of culturable microorganisms from USV missions ranged from 0 (LOD) to 42,411 CFU/m³ on R2A and from 0 (LOD) to 974 CFU/m³ on MOA, respectively (Table 1). PM concentrations from USV missions ranged from 2 to 8 μg/m³ for PM1, 3 to 24.1 μg/m³ for PM2.5, and 3 to 25.4 μg/m³ for PM10 (Table 1). The concentration of culturable microorganisms (bacteria and fungi) from UAS flights ranged from 6 to 9 CFU/m³ (over 10-min sampling intervals) (Table 2). Figure 4 shows PM concentrations for the six missions with outliers removed. Unexposed plates of R2A and MOA (controls for sUAS missions) did not yield any culturable microorganisms. Impinger control collections (i2 and i5) combined across missions for matching media types (R2B or MOB) did not yield any culturable microorganisms. Wind speed varied from 1 to 9.5 knots (0.5 to 5.0 m/s) at the Great Pond, Falmouth, MA, United States (Figure 5).

Six USV missions and four UAS missions were conducted on October 3, 2017 at Claytor Lake, Dublin, VA, United States (Tables 1, 2). The concentration of culturable microorganisms from USV missions ranged from 5,130 to 56,809 CFU/m³ on R2A (Table 1). No culturable microorganisms were observed on MOB (Table 1). PM concentrations from USV missions ranged from 0 (LOD) to 288 μg/m³ for PM1, from 2.5 to 18 μg/m³ for PM2.5, and from 2.6 to 19 μg/m³ for PM10 (Table 1). Figure 6 shows PM concentrations for the six missions (with outliers removed). The concentration of culturable microorganisms (bacteria and fungi) from UAS flights ranged from 12 to 16 CFU/m³ (over 10-min sampling intervals) (Table 2). Unexposed plates of R2A and MOA (controls for sUAS missions) did not yield any culturable microorganisms. Impinger control collections (i2 and i5) combined across missions for matching media types (R2B or MOB) did not yield any culturable microorganisms. Wind speed varied from 1.4 to 9.5 (0.7 to 5.0) knots during sampling at Claytor Lake, Dublin, VA, United States (Figure 5).

For the 1 μm beads, the peak of the aerosol size distribution in the chamber was 1.11 μm, matching the nominal size of the beads well. At this size, the impinger efficiency was 75.27 ± 8.26% (data not shown). For the 3 μm beads, the peak of the aerosol size distribution was 3.28 μm, matching the nominal size of the beads well. At this size, the impinger efficiency was 99.32 ± 0.590% (data not shown).

Comparisons between the reported concentration for the PMS7003 OPC (OPCA) and SDS021 OPC (OPCB) sensors to the TSI APS during trials are shown in Figure 7. High particle concentrations during laboratory testing may have contributed to observed differences in concentrations reported by the PM sensors compared to those reported by the APS.

The SDS0021 sensor at 0.1 m showed a significant correlation with wind speed reported by the USV mounted meteorological sensor at Claytor Lake, Dublin, VA, United States for PM2.5 (P < 0.001), PM2.5 and PM10 (P < 0.001). Correlations were also found at the Great Pond, Falmouth, MA, United States for wind speed with PM2.5 (P = 0.016) and PM10 (P = 0.006) for the PMS7003 senor at 0.1 m and PM2.5 (P < 0.001) and PM10 (P = 0.0316) for the SDS0021 sensor at 0.1 m. PM sensors for Claytor Lake, Dublin, VA, United States reported similar concentrations (P-values were <0.001), except the SDS0021 sensor at 0.1 m with both PMS7003 sensors at 0.1 and 1.1 m for PM2.5 and PM10 (P-values ranged from 0.299 to 0.551). For Great Pond, Falmouth, MA, United States concentrations were on average higher at 1.1 than 0.1 m for both of the PM sensors (OPCA and OPCB) across all PM sizes reported. Both PM sensors for the Great Pond, Falmouth, MA, United States reported similar concentrations (P-values were <0.001). For Claytor Lake, Dublin, VA, United States concentrations were on average lower at 1.1 than 0.1 m for both of the PM sensors (OPCA and OPCB) across all PM sizes reported. PM concentrations were highly correlated with wind direction for all sensors and PM sizes at Claytor Lake, Dublin, VA, United States (P < 0.001) and the Great Pond, Falmouth, MA, United States (P-values were <0.001).