Sea spray aerosols (SSAs) are generated over a large portion of the Earth's surface and form a large fraction of aerosol particulates present in the atmosphere (e.g., Lewis and Schwartz, 2004). They are critically important components in global biogeochemical cycles (e.g., Solomon et al., 2007) and important modifiers of atmospheric radiative budgets. Marine aerosols are generated primarily by processes associated with the formation of bubble plumes and foams generated by the actions of breaking surface waves. Breaking waves themselves play an important role in many additional processes at the air–sea interface including mixing, current formation, heat and momentum flux, and the bubbles entrained by breaking waves enhance gas transport, scavenge biological surfactants, and generate ambient noise in addition to creating aerosol particles (e.g., Woodcock, 1953; Wallace and Duce, 1978; Rapp and Melville, 1990; Tseng et al., 1992).

Oceanic whitecaps (which are the high optical albedo footprint of a breaking surface wave) typically form once wind speeds greater than approximately 3 ms-1 blow over a sea surface of sufficient fetch. Breaking itself includes the impaction of the overturning wave crest with the sea surface and subsequent entrainment and fragmentation of air into a plume of bubbles. The plume evolves over a timescale of seconds to a few tens of seconds due to buoyancy and turbulent flow forces acting on the entrained bubbles. The air / water mixture of the breaking wave crest and the bubbles that reach the sea surface after breaking form the high albedo patch characteristic of a whitecap. Surface bubbles and the dense aggregations of bubbles that create surface foams are the primary source of marine aerosols as the bubbles rupture and produce a spray of jet and fluid film droplets that are ejected into the atmosphere (Lewis and Schwartz, 2004).

In order to study marine aerosol production it is beneficial to have a standardized method of creating them that mimics the formation processes associated with marine foam in repeatable, controlled conditions in the laboratory. Several different methods have been used to generate surrogate marine aerosols within enclosed tanks including pressurized atomizers (Svenningsson et al., 2006; Riziq et al., 2007; Saul et al., 2006; McNeill et al., 2006; Braban et al., 2007; Niedermeier et al., 2008; Taketani et al., 2009), forcing air through glass filters or sintered materials (Cloke et al., 1991; Martensson et al., 2003; Sellegri et al., 2006; Keene et al., 2007; Tyree et al., 2007; Wise et al., 2009; Hultin et al., 2010; Fuentes et al., 2010) and by a plunging water jet (Cipriano and Blanchard, 1981; Sellegri et al., 2006; Facchini et al., 2008; Fuentes et al., 2010). The detailed investigations by Sellegri et al. (2006) and Fuentes et al. (2010) have shown that the best method for the generation of proxy marine aerosols is by creating a bubble plume from a plunging jet of water. In addition Collins et al. (2014) have shown that the method of bubble production influences the chemical composition of laboratory-generated sea spray aerosol, with a plunging water method showing better agreement with aerosol produced from laboratory breaking waves than did aerosol generated via the sintered glass filter method. The plunging jet apparatus used by Fuentes et al. (2010) used a relatively small volume of water (6 L) in an 11 L tank filled to a depth of 11 cm. Using a modification of the prior plunging water techniques, Stokes et al. (2013) developed the Marine Aerosol Reference Tank (MART) system that accurately reproduced the bubble plumes and marine aerosols characteristic of an oceanic whitecap. By using an intermittent plunging sheet of water in a larger (210 L) tank bubble plumes are formed that mimic the oceanic bubble size distribution, including critical bubbles larger than the Hinze scale (the transition point between bubbles stabilized by surface tension and bubbles subject to fragmentation by turbulence at approximately 1 mm scale), and have a temporal evolution similar to plumes measured in the ocean and in large laboratory wave tanks.

The two primary production mechanisms of sea spray aerosols at moderate wind speeds are the disintegration of the thin fluid films associated with whitecap foam (film drops) and the breakup of the jet of water formed at the base of a bubble shortly after the rupture of its film (jet drops). Both of these mechanisms are known to be sensitive to bubble size. It follows that an essential requirement of any laboratory system designed to produce nascent SSA is the reproduction of the numbers and sizes of bubbles entrained by breaking waves in the open ocean. Few bubble size distributions from natural breaking waves have been acquired because of the difficulty of making measurements in stormy conditions and other natural hazards (Herrero, 1985; Melville, 1996; de Leeuw and Cohen, 2002; Stokes et al., 2002). However, some oceanic measurements are available as well as a number of laboratory studies (e.g., Monahan and Zeitlow, 1969; Cipiriano and Blanchard, 1981; Bezzabotnov et al., 1986; Lamarre and Melville, 1994; Loewen et al., 1995; Leighton et al., 1996; Deane and Stokes, 2002; de Leeuw and Cohen, 2002; de Leeuw and Leifer, 2002; Leifer and de Leeuw, 2002, 2006; Stokes et al., 2002) and are summarized in Fig. 1 of Stokes et al. (2013). It is now known that there is a scale dependence to the bubble creation physics, differentiated by a length scale known as the Hinze scale (Deane and Stokes, 2002). The Hinze scale (aH) defines the radius of a bubble for which surface tension forces, which tend to keep bubbles spherical, are disrupted by distorting pressure fluctuations associated with fluid turbulence. This scale is of the order of 1 mm in spilling and breaking waves. Bubbles smaller than the Hinze scale are stabilized to fragmentation by fluid turbulence, whereas bubbles larger than this scale are subject to a turbulent fragmentation cascade.

The power law dependence of the bubble size distribution as a function of bubble radius is also different for bubbles smaller and larger than the Hinze scale. Smaller bubbles have a somewhat variable power law scaling, a-n with n taking values between approximately 1 to 2. The physics of bubble fragmentation and bubble degassing drives a steeper power law dependence for bubbles larger than the Hinze scale with n taking values between approximately 3 to 4 (Fig. 1). Important points are that (1) breaking oceanic whitecaps can produce large bubbles, greater than 1 mm radius and up to 4 mm radius (Bowyer, 2001), and (2) the power law scaling of the generation of these bubbles is controlled by fluid turbulence within the whitecap and differentiated by the Hinze scale. In order to accurately reproduce nascent SSA, the laboratory bubble generation mechanism needs to produce bubbles larger than the Hinze scale and reproduce the power law dependence those bubbles acquire through fragmentation in fluid turbulence.

The original MART system was constructed to closely mimic the bubble plume, foam, and aerosol generating mechanisms active during oceanic wave breaking and to provide a portable, controllable environment in which to explore and sample these processes (Stokes et al., 2013). The primary design of MART included a flow-controlled closed-loop circulation system that draws water from the tank bottom, a tank-top spillway or waterfall to produce a plunging sheet that impacts the water surface within the tank to produce a bubble plume, and an air-tight headspace for controlled aerosol sampling while the system is operating. By varying the temperature of the tank contents, the water chemistry and the characteristics of the plunging sheet (volume, angle and distance of drop, timing of the intermittency) a wide range of experimental conditions can be realized. The tank itself can also be used as an incubator for the growth of planktonic organisms to investigate the influence of biogenic exudates on SSA formation (Lee et al., 2015). A limitation with the MART system is that it can be difficult to culture delicate organisms in the reservoir while the external circulation pump (1/3 HP centrifugal pump) is operational because the high flow rates (70 L min-1 within the pump casing and up to 15 L min-1 in the waterfall flow) create high levels of fluid shear that is damaging to fragile cells. Hence, when including cultured cells in the experimental system it is necessary to limit pump cycling (and aerosol generation) to after the culture has reached its exponential growth phase or reached a cell density where losses due to pump cycling do not exceed cell creation rates.

The miniMART system (Fig. 2) described here was designed to provide a gentle method of plunging jet generation that would minimize destructive shear on cultured organisms and still permit the continuous generation of aerosols for study. It was fabricated using components that are readily available and constructed of stainless steel, Plexiglas and silicone wherever possible to minimize chemical contaminants and facilitate cleaning. The main tank (25 × 25 × 30 cm, 19 L total volume) was made from 1.5 cm thick Plexiglas with an O-ring sealed, 20 mm thick Plexiglas lid to provide airtight integrity. Separate ports are available for sampling both the atmospheric headspace and subsurface water in the tank. Port sampling tubes are made from stainless steel and positioned within 1 cm of the water surface to minimize the effects of particle losses to the Plexiglas tank walls.

Inside the tank, a 20 cm diameter, 8 cm deep, compartmentalized water wheel, and fabricated after an ancient sakia design, is rotated at approximately 8 RPM by an externally mounted 1/15 HP motor attached to a shaft-sealed axle that penetrates the tank rear wall. A thin silicone gasket was used to seal the compartmentalized wheel to its removable lid, providing access to its internal surfaces for cleaning. The contact between the silicone and the water is minimized to a hairline gap between the lid and wheel once assembled, to mitigate potential effects of the silicone on water surface microlayer chemistry. The two internal chambers of the wheel provide the intermittent release of a 70 mL water jet from approximately 10 cm above the water surface within the tank (when filled with approximately 6 L of water) via a hole in the chamber wall. The plunging jet sweeps across the water surface when a chamber crosses the apex of its rotation (maximum height above the water surface) while the opposite chamber is synchronously filling beneath the water surface. The plunging jet impacts the water surface and produces a bubble plume that mimics the plunging jet of water from a breaking wave crest without the need for a powerful external pump. The small size of the miniMART system allows it to be partially submerged (to the water fill line) in a temperature-controlled bath allowing stabilization of its internal temperature if necessary for plankton culture growth. Alternatively, the miniMART can be operated in a temperature-controlled incubator or room, a technique that has proven viable even for the much larger MART system.

Before experimentation the miniMART system is cleaned to minimize contamination. The internal surfaces are scrubbed with 100 % percent isopropanol and then the entire system is filled and the sakia wheel circulated with a 10 % isopropanol / deionized water solution for approximately 30 min. After circulation the tank is drained and then rinsed and filled with deionized water, and the system again circulated. Lastly, the system is flushed with filtered freshwater or seawater for experimentation. The system is considered clean when measurements of surface tension from water samples are the same as those from the filtered water supply used for experimentation (approximately 72 mNm-1 at room temperature measured using the Wilhelmy plate method with a Krüss K3 tensiometer.)

As noted by Sellegri et al. (2006) and Fuentes et al. (2010), a plunging water jet best replicates the bubble plumes generated by an oceanic whitecap. Comparison of the bubble plume formed by the miniMART system to those generated by air flow through sintered glass filters and to those formed in oceanic waves and within the larger MART system (Fig. 1) illustrates that a plunging sheet of water forms a broader spectrum of bubble sizes than the sintered glass filters tested, including bubbles larger than about 1 mm in radius. The slopes of the bubble density size spectrum in the miniMART plumes are very similar to the slopes of oceanic and laboratory breaking waves at sizes smaller and larger than the Hinze scale (aH) as well as to the larger MART system. For comparison, the bubble plumes generated by sintered filters have a much narrower size spectrum and tend not to include bubbles larger than about 800 µm radius.

The bubble plumes generated by the plunging jet within miniMART penetrate approximately 15 cm beneath the water surface, which is not as deep as the plumes generated by MART or by spilling breakers in the lab and ocean (Deane and Stokes, 2002). However, the intermittent cycling of the plunging jet in miniMART system allows the bubble plume and resulting surface foam patch to evolve over time, creating a bubble and aerosol source that seems to be a fairly close match to the decaying patches of foam produced by whitecaps than that provided by constant, stationary jets. The importance of decaying foams (as opposed to pseudo-steady-state foams, for which decay rates are matched by bubble entrainment rates) remains an open question, but may be important. For example, the jet drop production mechanism may be somewhat suppressed in steady-state foams if they are more than a single bubble layer thick because the top layer of foam film can absorb jet drop aerosols produced at the air–water interface (Collins et al., 2014). Foams allowed to decay, even if they are initially three-dimensional in structure, will eventually devolve into two-dimensional rafts of bubbles which will not suppress jet drops.

The particle number distribution measured using the miniMART and MART system are similar to the size distribution obtained by Fuentes et al. (2010). It is notable that the particle number distributions obtained using the miniMART (Fig. 4) and MART systems have less pronounced characteristics of sub-100 nm modes, with the dominant number distribution mode around 200 nm, broadly tailing off to both larger and smaller sizes (see details for MART in Fig. 5, of Stokes et al., 2013). This result is consistent with the broad bubble size spectrum and accurate representation of bubbles larger than 1 mm that is achieved by both miniMART and the larger MART system. Particle number distributions measured in both are in strong agreement with those previously measured from breaking waves in the Scripps Institution of Oceanography Hydraulics Laboratory (Prather et al., 2013). These measurements highlight the importance of an accurate representation of bubble formation processes in the creation of sea-spray aerosol in the laboratory. The primary difference between the miniMART and MART systems is the lower particle flux generated by the smaller and less energetic plunging jet in miniMART. For example, the submicron- and supermicron-sized particle number, surface area and mass concentration in MART were approximately 5000 and 345 cm-3, 1260 and 2800 µm2 cm-3, and 200 and 1735 µg m-3 (assuming a particle density of 1.8 g cm-3), respectively, at a flow rate of 3 slpm (Stokes et al., 2013). While for the miniMART, these numbers were approximately 90 and 60 cm-3, 160 and 900 µm2 cm-3, 50 and 125 µg m-3, respectively, at a flow rate of 2.6 slpm, necessitating longer sample integration times for some instrumentation, like the SMPS.

The reduced particle number concentrations in miniMART, in comparison to MART, can present a challenge for particle instrumentation (e.g., size resolved cloud condensation nuclei measurements). For instruments where the noise is dominated by counting statistics, signal-to-noise ratios can theoretically be improved by signal averaging. An important consideration, with respect to miniMART, is the stability of the particle source and air delivery as a function of instrument integration time. Allan variance can be used to determine the timescale for which signal averaging in the miniMART will no longer improve instrument signal-to-noise ratio (Werle et al., 1993). Twelve continuous hours of 1 s CPC measurements from a miniMART containing a 500 mM NaCl solution are shown in Fig. 6a. The Allan variance was calculated from these data and are shown in Fig. 6b. The analysis indicates that improvement in signal-to-noise ratio will be achieved for averaging times up to 100 s, after which further signal averaging will result in a decrease in the signal-to-noise ratio. Further work is required to establish the experimental factors that control this optimum averaging time.

A primary motivation for the fabrication of miniMART was to create an SSA analogue that allowed continuous aerosol sampling during the growth and culturing of planktonic cells. Figure 7 shows data collected during a 12-day miniMART incubation of sand-filtered seawater spiked with nutrients at 0 h. Aerosols were sampled continuously with an APS from the tank headspace with a carrier gas flow of 1.9 slpm. In addition, chlorophyll a concentrations and dissolved organic matter (cDOM) concentrations were measured at semi-regular intervals from miniMART water drawn via a peristaltic pump into a closed-loop analysis system (Wetlabs Ecotriplet), and then returned, to prevent the loss of water from the system during sampling. Exponential growth of microorganisms (primarily diatoms) peaks around day 4 with an increase in the number density of aerosols increasing after the initial bloom and while the chlorophyll a concentrations drop, associated with the death of the diatoms and rapid increase in the number of bacteria and viruses which cause cellular lysis and the increase in dissolved organics. Similar preliminary experiments have been run showing multiple microbial blooms and crashes during miniMART incubations for weeks in duration. Understanding the factors that drive the variability in the produced SSA particle concentrations that is evident in Fig. 7 is the subject of future work.

In order to mimic the SSA created by oceanic whitecaps any surrogate system must reproduce the complex two-phase flows, bubble plumes and surface foam patches naturally generated during a breaking wave. These conditions can be accurately replicated in large seawater breaking wave channels. However, these facilities are not readily available, and due to their extremely large volume it is extremely difficult to enclose them for high fidelity aerosol sampling and difficult to carefully control the environmental conditions to allow replicate experiments. Sintered glass filters (frits) bubbling air in an enclosed container produce controllable plumes; however, the bubbles produced are constrained to a narrow size spectrum much more narrow than that observed in a natural whitecap.

When using plunging water to create bubble plumes, it is important that the falling sheet or jet has the appropriate scale of surface roughness before impacting the water surface in order to create the correctly sized voids along the air–water interface (Zhu et al., 2000). The larger voids are important for producing the correct plume bubble size distribution that includes bubbles larger than the Hinze scale. Stationary, narrow cross-sectional area and high velocity jets may not entrain large bubbles characteristic of whitecaps without the correct scale of disturbances on their surface before impacting the water.

It is apparently important that any bubble plume surrogate provide the correct intermittency in production. Natural whitecap plumes and the resulting surface foam evolve over a timescale of seconds to tens of seconds, whereas continuous water jets impacting the surface at a fixed location create subsurface flow fields unlike breaking events. Continuous sparging of air through frits and nozzles or air entrainment by continuous jets can also create three-dimensional surface foams that do not evolve and dissipate like those within oceanic whitecaps, and these can bias physical and chemical attributes of the aerosols created when the bubbles rupture (Prather et al., 2013; Collins et al., 2014).

The bubble plume and resulting aerosol particle size distribution generated within the miniMART and MART systems resembles that generated from breaking waves within the SIO glass-walled wave channel. Confining the bubble generation to a smaller headspace air volume (< 50 L in the MART and ∼ 10 L in miniMART) as compared to the wave channel permits a significant increase in particle number concentrations (from 100, to > 5000, to approximately 500 particles cm-3, for the wave channel, MART and miniMART, respectively). As a result, the surrogate MART and miniMART systems enable a wide variety of measurements (e.g., size resolved hygroscopicity and heterogeneous reactivity) that are not feasible at the low number concentrations produced in the wave channel and allow for the controlled study of the chemistry and physics of marine bubbles, foam and aerosols. In these systems, experiments are more easily repeatable even while environmental variables, like the seawater and atmospheric chemistry and the physical forcing mechanisms controlling the plume dynamics, are manipulated. The pump-free action of miniMART allows the long-term growth and monitoring of delicate planktonic cell cultures while continuously producing aerosols for study, with the caveat that the flux of particles is less than in the larger MART system, a factor which must be considered in any particle sampling protocols. For experiments requiring the generation and collection of large numbers of aerosol particles that do not require the continuous presence of delicate organisms in the water, the original MART system remains the breaking wave surrogate of choice.

All data sets used in this research can be provided through contact with the corresponding author. In addition, access can be acquired via contact with the NSF Center for Aerosol Impacts and the Environment (CAICE) at UCSD (http://caice.ucsd.edu/index.php/research/tools/).