Wind-driven spume droplet production and the transport of Pseudomonas syringae from aquatic environments

Flume design for imaging experiments (Flume A)

The flume for imaging experiments (Flume A) was constructed from 0.635 cm width Plexiglas with inner dimensions of 90.2 cm ×3.2 cm ×34.3 cm, as shown in Fig. 2A. The flume was a closed system with one 1.27 cm diameter inlet pipe reaching to the surface of the water and one 1.27 cm diameter outlet pipe exiting near the water surface. The inlet pipe had a nozzle shape designed to direct water away from the side of the flume minimizing droplets landing on the flume side. The flume was filled with deionized water to a height of 9.8 cm. Experiments were conducted at ambient room temperature. The flume outlet was connected to a vacuum that pulled air across the water surface simulating wind moving on the water surface. The vacuum was controlled with a 5 Amp FS-5F single pole 120 V rotary dimmer switch (Lutron, Coopersburg, PA, USA). A 29 series II multimeter (Everett, WA, USA) with a resolution of 0.1 V and accuracy of ±1% was used to measure the voltage of the dimmer switch. A EA-3010 anemometer (LaCrosse Technology, La Crosse, WI, USA) with a resolution of 0.1 m/s was used to measure wind speed. The wind speed U was measured at z = 2.5 cm above the water surface and on the downwind edge of the camera’s field of view. Measured wind speeds were nearly identical (and thus stable) up to a distance of 40 cm downwind from the inlet nozzle (well beyond the viewing window of the camera) for flume flow rates of 3.0 m/s. At distances greater than 40 cm from the inlet nozzle, the windspeed increased up to a factor of three at the outlet (located 60 cm downwind from the inlet). Consequently, the flume demonstrated spatially uniform (or non-accelerating) flow of near constant velocity near the water surface, within the viewing window of where the wind-driven spume droplets were observed by the camera. The outlet valve to the vacuum was adjusted until the anemometer fluctuated by ±0.2 m/s of the target speed. After an initial calibration was made, the wind speed was controlled only with the multimeter. A SpectroLED-14 light (Genray, New York, NY, USA) was placed behind the flume. A Photron FASTCAM Mini UX100 camera (Photron, San Diego, CA, USA) with a micro-Nikkor 105 mm f/28 lens (Nikon, New York, NY, USA) was used to record video data at 6250 frames per second (0.16 ms between frames) in 0.9 s increments with a resolution of 1,280 × 800 pixels. The camera field of view was 6.6 cm of length by 4.1 cm of height. The resolution was 52 µm per pixel. The depth of field allowed the entire 3.8 cm width of the flume to be in focus. Videos were captured at four wind speeds (U = 3.5 m/s, 4.0 m/s, 4.5 m/s and 5.0 m/s). At 3.5 m/s, 12 video increments were recorded on 2 days (n = 122). At 4.0 m/s, 12 video increments were recorded on 4 days (n = 563). At 4.5, m/s four video increments were recorded on 2 days (n = 672). At 5.0, m/s 3 video increments were recorded on 2 days (n = 771). Assuming a power-law for the wind profile (Etkin, 1981), we have U₁₀∕U = (z₁₀∕z)^α where U is the wind speed measured at height z (in m) above the water surface, and U₁₀ is the wind speed at the reference height z₁₀ = 10 m. While there is significant uncertainty in the exponent α, which depends on atmospheric stability, surface roughness and nearby obstacles, we use a value of 0.1 which is appropriate for near-neutral conditions over a large body of water (Hsu, 1988). We did not account for any potential wall effects of our narrow flume (Williams, 1970).

Flume design for transport experiments for aqueous suspensions of P. syringae (Flume B)

The flume initially designed for the imaging experiments was modified (Flume B) to allow for a series of transport experiments with aqueous suspensions of P. syringae. Inner dimensions of Flume B were 90.2 cm × 3.2 cm × 34.3 cm. A droplet-sampling trap (v-shaped with a horizontal inlet) was used to capture cells of P. syringae from an aqueous suspension in the flume. The orifice of the trap had a diameter of 1.27 cm, and the inlet was 4 cm long, measured from the center of the trap (Fig. 2B). The trap was designed to be 56 cm tall, to prevent droplets from reaching the top end due to precipitation and deposition at the walls. When the trap was installed for experiments with P. syringae, additional turbulence was introduced in the system. Therefore, all wind speed measurements for the P. syringae experiments were conducted without the trap installed. At the top of the trap tube, a flexible hose of 0.64 cm diameter connected the tube to an impinger filled with 20 ml of 0.2 µm filter sterilized water. A vacuum was used to pull air through the impinger, and was controlled by a valve on the vacuum line to achieve a steady flow rate. The trap system was surface sterilized with 70% ethanol prior to each experiment. Two different trap extension lengths allowed sampling at two different heights, 5 and 10 cm above the suspension of P. syringae in the flume. All parts of the trap system were rinsed with 1 ml wash of sterile distilled water to test for any bacteria deposited on the different pieces of the trap.

Image analysis with Flume A

Video images were converted to black and white. The centroid and the centroid trajectory of each droplet were found across a stack of images. Diameters of the droplets were determined from the pixel area of the centroid data using the mean of the first ten images to minimize error due to small variations. Variation in the estimated diameter across images was typically ±5%, increasing to ±15% for the smallest resolved diameters (52 µm). Ejection speed was determined from the droplet centroid displacement between images, as near as possible to the point of ejection from the water surface. Variation in the estimated speed, determined by the standard deviation in speed estimates across ten images, was ±5%. The angle of the droplets was determined from a line tangent to the droplet trajectory as near as possible to the point of ejection from the water surface, to reduce error due to droplet movement being affected by the wind. Variation in the estimated angle, based on limits of resolution, was ±2 degrees. Droplets were categorized according to the mechanism of droplet production according to visual inspection of the videos, as in Fig. 1: bubble bursting, fragmentation, or unknown. Droplets were classified as ‘unknown’ if the moment the droplet first appears was not caught in the field of view. At U = 3.5 and 4.0 m/s, the populations of droplets from the bubble bursting and fragmentation mechanisms were large enough (n > 20 in each group) for the two populations to be compared.

Gamma distribution fit

Gamma distribution for diameters D, is given by the probability distribution function, ${\displaystyle p\left(D\right)=\frac{{\left(\frac{D}{{\theta }_D}\right)}^{k_D-1}{e}^{-\left(\frac{D}{{\theta }_D}\right)}}{{\theta }_D\Gamma \left(k_D\right)}}$ where the shape variable is $k_D={〈D〉}^2∕{\sigma }_D^2$, the scale variable is ${\theta }_D={\sigma }_D^2∕〈D〉$, and $\Gamma \left(\cdot \right)$ is the gamma function. This gamma distribution has mean $〈D〉$ and variance ${\sigma }_D^2$. As a pre-processing step to remove significant outliers, we binned the diameters and considered only diameters up to a cutoff diameter, the minimum diameter bin with a number of elements less than 0.02n_max, where n_max is the number of elements in the bin with the maximum number of elements. The same procedure was followed for the gamma distribution of velocity, yielding the shape variable $k_V={〈V〉}^2∕{\sigma }_V^2$ and the scale variable ${\theta }_V={\sigma }_V^2∕〈V〉$.

Experiments with P. syringae in Flume B

A 3L suspension of P. syringae (ice+ BAV strain #892 (Pietsch et al., 2015)) was diluted from a 4 mL liquid culture grown overnight (12–14 h) at 22 °C in R2 broth (3.15 g/L R2 broth, TEKnova #R0005, Hollister, CA 95023) (Reasoner & Geldreich, 1985). This starting culture was first diluted to an optical density of 0.2 at 600 nm. Control R2A plates were spread with 0.020 mL of a 1 × 10⁻⁶ dilution of the starter culture to confirm CFU in a countable range of 20–100 colonies per plate. For each flume experiment, a starter culture (optical density of 0.2 at 600 nm, corresponding to a range of 1,000 to 5,000 CFU per mL) was diluted in 10 mM MgSO₄ to make 3 L of an aqueous suspension of P. syringae which was added to the flume. This aqueous suspension (flume solution) was used to fill the flume to a height of 9.66 cm, corresponding to a water volume V_Psyringae = 2.7 L. Growth of P. syringae was determined by CFU counts on two types of agar plate, R2A (R2 broth made with 15% agar (Fisher 9002-18-0)) and KBC (Kings B medium with 15 g/L proteose peptone, 1.5 g/L anhydrous K₂HPO₄, 10 mL/L 100% glycerol, 6 mM MgSO₄) with 24 mM H₃BO₃, cephalexin (10 mg/L), and cycloheximide (50 mg/L) (Mohan & Schaad, 1987)). Samples were plated in triplicate and colonies were counted and reported as mean CFU per mL plated. Experiments were conducted with aqueous suspensions of P. syringae to determine the number of cells being transported at different heights (5 and 10 cm) under different wind speeds (3.5 and 4 m/s).

Experimental parameters for height and wind speed were held constant for each experiment with a trap collection time of 30 min (1,800 s). Trap collections were sequentially carried out in triplicate. For each 30-minute trap collection, 20 mL of sterile water was loaded into the 50 mL impinger apparatus. Prior to trap collections, 0.10 mL of the flume solution was plated in triplicate on R2A and KBC to obtain tank start values for P. syringae. Tank start values were determined for each experiment. Following trap collections, the same plating scheme was carried out for Aerosol Experiments 1, 2, and 4. The end tank aliquot for Aerosol Experiment 3 was not collected. For each 30-minute trap collection, volumes were recovered and 0.10 mL plated in triplicate on R2A and KBC plates. For collected trap volumes less than 2.0 mL, an aliquot of 0.50 mL sterile water was used to rinse the trap. This volume change was calculated as a dilution when determining CFU/mL values for the trap collections in Aerosol Experiment 2 and 4. For each experiment, the three impinger collections were combined and concentrated by filtration through a 0.2 µm filter, followed by an immediate resuspension in 5 mL of sterile filtrate via a 10-minute filter spin in a sterile 100 mL bottle. The combined impinged resuspension was plated in 0.20 mL aliquots in triplicate on R2A and KBC plates.

Rate of droplet production

Rate of droplet production, F_d, in m⁻² s⁻¹ of water surface was related to wind speed, U in m s⁻¹, measured at z = 2.5 cm above the water and related to equivalent wind speeds, U₁₀, at 10 m above the water surface (Fig. 3A and Table 1). The data fit a second order polynomial, $F_d=\kappa {\left(U-U_c\right)}^2$, for constant parameters κ in s m⁻⁴ and U_c determined from a non-linear curve fit, with an R² = 0.97 (it should be noted, however, that this is based on only four windspeeds tested in our study). A critical threshold wind speed of U_c = 3.15 m/s at 2.5 cm height (or equivalently U_10c = 5.7 m/s at 10 m height) was determined, below which there is no droplet production.

Distributions of droplet diameter, speed, and angle

Average and variance of droplet diameter, D, increased with wind speed; more small and large droplets are produced as well as the largest drop size (D_max ≈ 5.0 mm for U = 3.5 m/s and D_max ≈ 8.8 mm for U = 5.0 m/s). Probability distributions were best fit with a gamma distribution (Fig. 4 and Table 2), calculated from the average $〈D〉$ and variance ${\sigma }_D^2=〈{\left(D-〈D〉\right)}^2〉$ which provide the shape parameter, k_D, and scale parameter, θ_D. A log normal distribution (not shown) was also fit to the data (Kolmogorov, 1962; Oboukhov, 1962; Novikov & Dommermuth, 1997), but the gamma distribution was a better representation of the experimental data, giving lower error (having a higher R²) (Beck & Cohen, 2003; Villermaux, Marmottant & Duplat, 2004). Average droplet diameter and variance generally increased with wind speed. Probability distributions for droplet ejection speed, V, at each wind speed also showed a good fit to a gamma distribution (Fig. 5 and Table 2). Probability distributions for the droplet angle for the four wind speeds are given in polar plot form in Fig. 5, where zero degrees is the downwind direction. We see a wider distribution in angle for U = 3.5 and 4.0 m/s compared to U = 4.5 and 5.0 m/s.

Distributions of diameter, speed, and angle for bubble bursting and fragmentation droplets

Bubble bursting droplets were observed for 21% (26/122), 10% (285/563), 0.3% (2/672), and 0% (0 out of 771) of the droplets observed at U = 3.5, 4, 4.5 and 5 m/s, respectively.  Droplet data at the lowest wind speeds, 3.5 and 4.0 m/s were presented in categories of bubble bursting or fragmentation based on the observed mechanism of droplet production (Figs. 1A–1B). Diameter and speed of bubble bursting and fragmentation droplets indicated, in general, that fragmentation droplets had a wider distribution while the bubble bursting droplets had a narrower distribution, centered around smaller diameters (Figs. 1C–1D). Polar plots for the angle distributions of bubble bursting and fragmentation droplets indicated the bubble bursting droplets in general were between 20° and 100°, while the fragmentation droplets were between −40° and 70°. We were not able to determine the mechanism of production for a majority of the droplets at higher wind speeds, due in part to the noise of droplet splatter on the flume surface. We observed fewer bubble bursting droplets at higher wind speeds, but this may be due to the inability to distinguish bubbles in our experimental set up at higher speeds.

Droplet production mass flux

Mass flux ranged from 8.7  × 10⁻⁵ gm⁻²s⁻¹ for droplets with diameters D < 100 µm at U = 3.5 m/s to 200 gm⁻¹s⁻² for droplets with D > 1,000 µm at U = 5.0 m/s. Mass flux of droplets in each size range (D < 100 µm, 100 µm <D< 200 µm, 200 µm <D<1,000 µm) increased two orders of magnitude with wind speed from 3.5 to 5.0 m/s (Fig. 3B). Mass flux calculated from the gamma distribution showed a good fit with the mass calculated from the experimental data, over the range of diameters where comparisons could be made, indicating that the gamma distribution may be used to extrapolate the mass flux of droplets smaller than the experiment could resolve (D < 52 µm).

Droplet production momentum flux

Given their potential role in atmosphere-water momentum exchange (Veron et al., 2012), we report the momentum flux for small droplets (those with the potential to be suspended) as a function of both wind speed and angle (Fig. 6). Values of total momentum flux ranged from 1.4  × 10⁻⁴ gm⁻¹s⁻² for droplets with diameters D < 100 µm at U = 3.5 m/s to 4.6  × 10⁻² gm⁻¹s⁻² for droplets with diameters 100 µm <D < 200 µm at U = 5.0 m/s.

Observed and estimated transport of P. syringae at different heights under different wind speeds

The start (beginning of each experiment, prior to flume operation) and the end (end of experiment, after flume was turned off) concentrations are reported in Table 3. At a wind speed of about 3.5 m/s, aqueous suspensions of P. syringae were collected at rates of 283 cells m⁻² s⁻¹ at 5 cm above the water surface, and at 14 cells m⁻² s⁻¹ at 10 cm above the water surface (Table 3). At a wind speed of about 4.0 m/s, aqueous suspensions of P. syringae were collected at rates of 509 cells m⁻² s⁻¹ at 5 cm above the water surface, and at 81 cells m⁻² s⁻¹ at 10 cm above the water surface (Table 3). These trends correspond well to the respective volumes trapped at different wind speeds and at different heights. Increasing the height of the trap (from 5 cm to 10 cm), decreased the trapped volume by a factor of about 11. This is independent from the wind speed. Increasing the wind speed from 3.5 to 4.0 m/s increased the trapped volume by a factor of 3–3.2, which is independent of the height. P. syringae were not found above the trap in the flexible hose.

We estimated the flux of cells emerging out of the tank water surface by multiplying the total mass production flux by the measured tank cell density. At the 3.5 m/s wind speed, the total mass production flux is 3.4 gm⁻² s⁻¹ and at 4.0 m/s, it is 4.7 gm⁻² s⁻¹. Droplets with diameters exceeding 1000 µm dominate mass production flux at both wind speeds. Based on these mass production flux values and the measured cell density in the tank at the start and end, we can estimate the cell surface flux (Table 3). During the duration of the experiment, the cell density in the tank typically increased by a factor of 2. At the 3.5 m/s wind speed, the estimated surface flux is higher than the measured 5 cm flux by a factor of 2.5–6.1, and higher than the 10 cm flux by a factor of 93–170. At the 4.0 m/s wind speed, the estimated surface flux is higher than the measured 5 cm flux by a factor of 1.5–3.3, and higher than the 10 cm flux by a factor of 26–52.