On global and regional scales, air-water gas exchange is an important process that strongly influences the budgets of biogeochemical trace gases, as well as the transport of volatile pollutants between the atmosphere and a water body. For these reasons, it is crucial to understand the mechanisms that control the rate of air-water gas exchange.
 Near-surface turbulence is presumed to be the dominant mechanism regulating the transfer velocity, k, of slightly soluble gases across the air-water interface in the absence of bubbles. The magnitude of k is determined by diffusion through a spatially and temporally varying boundary layer, the thickness of which is a function of near-surface turbulence and diffusion.
 Considerable effort has been spent on determining empirical relationships between k and wind speed [e.g., Liss and Merlivat, 1986; Nightingale et al., 2000; Wanninkhof, 1992; Wanninkhof and McGillis, 1999], since wind speed is relatively easy to measure and plays a central role in the generation of turbulence through the transfer of momentum to waves and currents at the ocean surface. For a wind-driven system, turbulence is generated near the air-water interface through shear, buoyancy, or large- and micro-scale wave breaking, among other processes. There is increasing evidence that k may strongly correlate with air-water surface roughness or the mean square slope of short wind waves [e.g., Bock et al., 1999; Jähne et al., 1987]. In the presence of surface films, k may be significantly reduced at a given wind speed or wind stress [Asher and Pankow, 1986; Jähne et al., 1987, 1979, 1985]. Surfactants can significantly dampen waves at high wave numbers and thus affect wave slope spectra [Frew, 1997]. This change in the wave spectra is linked to reduced gas exchange [Bock et al., 1999]. Micro-scale wave breaking has been suggested as a dominant mechanism for gas exchange at low to moderate wind speeds over the ocean [Zappa et al., 2001] (see also C. J. Zappa et al., Microbreaking and the enhancement of air-water gas transfer velocities, submitted to Journal of Geophysical Research, 2004) (hereinafter referred to as Zappa et al., submitted manuscript, 2004) and may explain the observed correlation between k and surface roughness. Regardless of the specific processes and details of the underlying physics, models generally attempt to parameterize k based on the assumption that turbulence regulates the exchange.
 Laboratory experiments and preliminary field studies show that raindrops falling on a freshwater surface significantly enhance k. Rain-induced gas exchange systematically increases with the kinetic energy flux (KEF) to the water surface supplied by the raindrops [Ho et al., 2000, 1997]. The enhancement in k by rain is dominated by the production of turbulence and secondary motions, whereas rain-generated bubbles contribute 0–20% of the total gas exchange, depending on rain rate, raindrop size, and gas solubility [Ho et al., 2000]. These findings suggest a possible mechanism for increasing the rate of air-water gas exchange in quiescent environments with little wind-forcing.
 Until now, all experiments aimed at quantifying rain-induced gas exchange have been performed in freshwater. The main difference between rain falling on freshwater and salt water is the resulting stratification of the upper water column in the case of a saltwater body. Rain promotes density stratification that inhibits turbulent mixing, but might also generate density-driven circulation, which can contribute to water column turbulence. Near-surface shear associated with rain-induced density stratification may affect the turbulence in the surface aqueous boundary layer and the ability to predict k. Thus the behavior and properties of these processes need to be understood and measured for development of adequate models for rain-induced gas exchange.
 Because it is difficult to conduct controlled experiments in the open ocean, the Biosphere 2 ocean was chosen for pilot studies of rain-induced air-sea gas exchange. The facility is ideally suited for such an experiment because (1) the processes that are responsible for rain-induced gas exchange in the Biosphere 2 ocean should be similar to those in the real ocean; (2) most of the ceiling height above the Biosphere 2 ocean is greater than 13 m, which allows realistic rain to be generated (i.e., raindrops of reasonable size approaching terminal velocity); and (3) the presence of a wave generator offers the possibility of studying gas exchange driven by both the interaction of a wavy surface and the additional effect of rain.
 In the following, we present the results of an experiment conducted at the Biosphere 2 Center: Bio2 RainX II. The gas transfer velocity was determined by performing an SF6 evasion experiment. Raindrop size distributions were measured with a Rain Imaging System (RIS), records of high-resolution temperature and salinity gradients were made using the Skin Depth Experimental Profiler (SkinDeEP), surface wave fields were measured using a laser altimeter, 3-D current fields were measured using a bottom mounted acoustic Doppler current profiler (ADCP), turbulence measurements were performed with three acoustic Doppler velocimeters (ADV), and the thermal signature of water surface was measured using an infrared (IR) imager. Together, these measurements allow us to study the influence of rain on air-sea gas exchange, as well as to elucidate the mechanisms responsible for the observed effect.
 During Bio2 RainX II, three rain experiments were conducted. The first (RE1) was a short rain event (53 min), during which equipment was tested. The second (RE2) was a longer rain event (122 min), during which an SF6 evasion experiment was conducted. The third (RE3) was a very short event (15 min) during the night, when infrared measurements were made. The length of RE2 was dictated by the fresh water storage capacity of the reservoirs and by the range of tolerable water level in the ocean, as well as the salinity limit imposed by the requirements of organisms in the ocean.
2.1. Biosphere 2 Ocean
 The Biosphere 2 ocean contains 2650 m3 of saltwater with a nominal salinity of 35.5. It has a surface area of approximately 675 m2 and an average depth of 3.5 m. The ocean is maintained at a constant temperature of 26.5°C by pumps that circulate the ocean water through a heat exchanger. On one side of the ocean, there is a vacuum wave generator capable of creating sufficiently energetic waves to circulate water in the ocean and enough turbulence to enhance air-water gas exchange. The entire length of the wave wall is partitioned into five sections of enclosed air space above the water surface. A vacuum is applied to the chambers, and the water level inside rises. When the water reaches a specified height, the vacuum is relaxed to atmospheric pressure and the water level falls, causing water to rush out of the chambers, and creates waves that propagate toward the beach (Figure 1). Wave amplitude is regulated by the water level in the chambers, which in turn is controlled by the pumping rate of the vacuum, and the frequency of the wave is regulated by the timing of the valves that apply and relieve the vacuum. For a more detailed description of the Biosphere 2 ocean, see Atkinson et al. .
2.2. Rain Generator
 Two water reservoirs (36 and 57 m3) were used to supply water to the rain generation system, via 10 cm OD PVC pipes and two 15 horsepower multistage centrifugal pumps. The tanks contained groundwater that had been purified by reverse osmosis to remove nutrients and other contaminants.
 The system used for generating rain during Bio2 RainX II consisted of a series of eight 2.5-cm OD polyvinylchloride (PVC) pipes hanging off messenger wires strung across the ocean. Attached to each pipe were three rain heads, for a total of 24 rain heads. The rain heads were modified commercially available irrigation devices (Rain Bird Xeri-Bird 8 MultiOutlet Emission Device) with short pieces of 0.3-cm ID latex tubing secured to the eight emitter nozzles on each device. The heads were oriented with the tubing facing up (head inverted) with pipe stems descending from the overhead 2.5-cm OD PVC pipes. The flow regulators were removed from each device, allowing for a higher flow rate of 15 L min−1. The flexible tubing encouraged random dispersion of the drops, as well as a spectrum of drop sizes to be generated. The rain heads were located 13 m above the ocean, allowing raindrops to achieve terminal velocity before impacting the surface.
2.3. SF6 Evasion Experiments
 To assess the effect of rain on air-sea gas exchange, an SF6 tracer release experiment was conducted during RE2. A predetermined amount of SF6 dissolved in water (∼1.3 × 10−5 moles) was injected into the ocean using a 60-mL syringe 12 hours in advance of RE2. Thorium isotope [Atkinson et al., 1999] and SF6 measurements (Bio2 RainX I, unpublished data, 2001) have shown that the Biosphere 2 ocean mixing time is of the order of a few hours. SF6 measurements made during a pilot rain experiment showed that the tracer remained laterally well mixed in the Biosphere 2 ocean during the rain events (Bio2 RainX I, unpublished data, 2001). The transfer velocity for no rain conditions (R = 0) was determined by measuring the decrease in SF6 concentration as a function of time in the ocean in the absence of rain over a period of 20 hours. Immediately prior to RE2, samples for SF6 were taken to establish the initial concentration. The gas transfer velocity during RE2 was determined by measuring the decrease of SF6 in the water with corrections applied for dilution due to addition of SF6-free rainwater (see below).
2.4. Water Sample Collection and Measurement
 A specially designed sample profiler allowed water to be drawn from different depths in the ocean for salinity and SF6 measurements. The profiler consisted of a 5 × 60 × 60 cm polystyrene float, and 14 lengths of 0.3175 cm ID flexible vinyl tubing, secured at different depths relative to the float along a weighted line. The float kept the sampling ports at fixed depths relative to the water surface (2, 3, 4, 6, 9, 17, 30, 49, 69, 90, 134, 210, 393, 596 cm) as the water level rose during rain and as waves passed the sampling point. The profiler was secured in place at a fixed location in the ocean by attachment to a boom extending away from the edge of the ocean.
 The inlet of each tube was protected with a screen to prevent blockage of tubing by sediments and organic matter in the ocean. During a typical sampling procedure, the tubing was attached to a peristaltic pump, and water was drawn at a rate of ∼100 mL min−1, which filled a measurement cell attached to a conductivity probe on a YSI 6600 multiparameter sonde. Each salinity measurement was completed when its value reached steady state, and the measurement cell was allowed to drain before refilling with sample from the next depth. During and after the rain event, the depths were sampled out of sequence to ensure that measurements at a given depth did not affect adjacent depths. Any bubbles forming on the conductivity cell were removed by lightly agitating the sonde.
 After completion of the salinity measurement, a three-way valve at the top of the sampling line was closed to prevent drainage back down the tube, and the peristaltic pump was disconnected. A 50-mL glass syringe was connected to the three-way valve, the valve was opened and water was drawn slowly into a 50-mL glass syringe for SF6 measurement. During water sampling for SF6, extreme care was taken to prevent the occurrence of bubbles in the sampling line or in the syringe. No samples were kept for analysis when bubbles were seen in the tubing or in the syringe.
 SF6 analyses were conducted using a headspace method described by Wanninkhof et al. . Glass syringes were filled to 30 mL of water, and then a headspace of 20 mL was created with ultra-high purity (UHP; 99.999%) N2. The syringes were then shaken vigorously on a mechanical shaker for 3 min to equilibrate the water with the N2 in the headspace. The gas sample in the headspace was then pushed through a drying column of Mg(ClO4)2 and into a sample loop. Subsequently, the sample was injected into a gas chromatograph equipped with an electron capture detector (GC-ECD) using UHP N2 as carrier gas. SF6 was separated from other gases at room temperature with a Molecular Sieve 5A column.
 During the RE2, samples were collected every 20 min at a fixed site from 14 depths in the ocean. Sampling for SF6 and salinity continued for 24 hours after the rain stopped to examine how the ocean relaxes to an equilibrium state, and to estimate k in the absence of rain.
2.5. Tracer Dilution Models and Gas Exchange Calculations
 The observed decrease in total tracer concentration in the ocean during the experiments was caused by gas exchange at the air-water interface and by dilution of ocean water by SF6-free rain. In order to separate the gas exchange component from the dilution component, the following dilution correction has to be applied to the measured SF6 concentration Cm(z) in order to obtain a corrected SF6 concentration C(z),
where s0 is the initial salinity of the ocean before the rain and s(z) is the salinity measured concurrently with Cm(z).
 For each profile, the mean SF6 concentration in the ocean is determined by
where C(z) is the concentration of SF6 measured at a certain depth z, V(z) is the volume of the ocean in the depth range represented by z (Figure 2), and Vtot and ztot are the total volume and depth of the Biosphere 2 ocean, respectively. The gas flux across the air-water interface, F, is related to the total change in mean SF6 concentration in the ocean with time by
where h is the mean depth of the ocean. The gas transfer velocity, k, is described by
where Cw is the SF6 concentration in the water directly below the air-water interface, and αCa is the solubility equilibrium for SF6 in the water. If the water is well mixed with respect to tracer concentration, equations (3) and (4) are combined, assuming that Cw = C, and integrating over Δt to obtain
where Ci and Cf are the initial and final mean tracer concentrations in the ocean, respectively. If there is a significant vertical gradient in tracer concentration, then equations (3) and (4) are combined to yield [Ledwell, 1982]
Equation (6) was used to calculate a gas transfer velocity from the tracer data. SF6 in the Biosphere 2 atmosphere, Ca, was non-negligible and was determined by sampling with 50-mL glass syringes and measurement using the GC-ECD.
 The gas transfer velocity for SF6 was normalized to a Schmidt number (Sc) of 600, corresponding to values for CO2 at 20°C using the relationship
where and are the gas transfer velocity and the Schmidt number for SF6 (782 for our experiment), respectively. Sc is defined as the ratio of the kinematic viscosity of water, v, to the diffusion coefficient of gas in water, Dg. It has been shown in models and experiments that for a clean wavy water surface, in the absence of bubbles, n equals −1/2 [Brumley and Jirka, 1988; Jähne et al., 1984; Ledwell, 1984].
2.6. Rain Rate and Drop Size Distribution
 During Bio2 RainX II, the raindrop size distribution (DSD) was obtained from data provided by a Rain Imaging System (RIS) developed at NASA. RIS is an optical system consisting of an analog black and white video camera that is pointed at a halogen flood lamp. The 100–300 mm lens was adjusted so that the center of the focal volume is located at 2 m from the camera. The field of view is 32 × 24 mm, and the depth of field is 13.6 mm. The distance between the camera and the light source was 3 m, so there was no significant droplet splash from the instrument into the measurement volume. The RIS camera and light were located on the coral reef as shown in Figure 1, and RIS was connected to an image analysis personal computer located on the beach area. The video data were compressed and archived for later processing.
 Digital images were obtained at a rate of approximately 50 Hz. Each image consists of 640 × 240 pixels. The pixel size is 0.05 × 0.1 mm and is adequate to identify and characterize drops larger than 0.5 mm diameter. The halogen lamp provides sufficiently bright illumination so that the camera operates with the shutter speed of 1/100,000 s. The terminal velocity of the water drops increases with drop size, such that for 1- to 5-mm drops, it varies from 2 to 5 m s−1. Together, the shutter speed, frame rate, and image cross section yield independent images that have insignificant blurring due to motion [see Saylor et al., 2002]. Each water drop presents a distinctive pattern: A water drop within the focal volume appears as a disc with a hole; a drop just outside of the focal volume appears as a solid disc; and drops still farther away from the focal volume lack sufficient contrast to be analyzed.
 The images were processed using a four-step algorithm: (1) a pixel scaling procedure to account for non-uniform illumination, (2) a threshold process to distinguish dark areas from bright ones, (3) a pattern recognition process to select only drops within the measurement volume, and (4) a particle sizing measurement to obtain the equivalent drop diameter. Equivalent drop diameter is a sizing measure that relates the observed area to that of a circle. Video data were obtained with the RIS during the entire experiment and all the images were analyzed to compute the DSD. Computations by Craeye and Schlussel  of drop velocity versus dropping height for various drop sizes show that for a fall height of 10 m, drops as large as 5 mm are within a few percent of terminal velocity. At the Biosphere 2 facility, the 13-m fall height of the drops is sufficient for drops to approach terminal velocity. Thus a relationship developed by Lhermitte  was used as a realistic estimate of drop velocities v in the RIS measurement volume,
where D is drop diameter in mm and v has units of m s−1. Ho et al.  proposed that rain-induced air-water gas exchange is correlated to rain kinetic energy flux (KEF). KEF can be derived from a DSD according to
where ρ is the density of water and KEF has units J m−2 s−1.
 During Bio2 RainX II, the spatial distribution of the simulated rain was assessed using water volume measurements from twelve 10-cm-diameter buckets. The buckets sat on floats, which were each tethered by a 1-m line to one of two ropes that extended from the wave generator to the beach.
2.7. High-Resolution Temperature and Salinity Profiles
 SkinDeEP, an instrument designed to make high-resolution profiles within the upper 10 m of the ocean, was deployed during Bio2 RainX II to study the evolution of the cool freshwater lens both during and after the rain events. SkinDeEP operates autonomously by changing its density. Positive buoyancy is achieved by pumping air from inside the body of the profiler into an external neoprene inflatable sleeve. The instrument sinks when the sleeve is deflated by returning the air to the interior. The instrument is equipped with millimeter resolution temperature and conductivity sensors mounted some distance from the top end-cap. Data are recorded only during the ascending phase of the profile when sensors are out of the wake of the instrument. For a more thorough description of the instrument, see Ward et al. .
 During the experiment, SkinDeEP was deployed at the deep end of the ocean (6.5 m), adjacent to the wave generator (Figure 1
Berger R, Libby WF (1969) Equilibration of atmospheric carbon dioxide with seawater: possible enzymatic control of the rate. Science 164:1395–1397.CrossRefGoogle Scholar
Broecker HC, Petermann J, Siems W (1978) The influence of wind on CO2 exchange in a wind-wave tunnel, including the effects of monolayers. J. Mar Res 36:595–610.Google Scholar
Broecker, W S, Peng T-H, Mathieu G, Hesslein R, Torgeson T (1980) Gas exchange measurements in natural systems. Radiocarbon 22:676–683.Google Scholar
Broecker WS, Peng T-H, Ostlund G, Stuiver M (1985) The distribution of bomb radiocarbon in the ocean. J Geophys Res 99:6953–6970.CrossRefGoogle Scholar
Cember, R (1989) Bomb radiocarbon in the red sea: a medium-scale gas exchange experiment J Geophys Res, 94:2111–2123.CrossRefGoogle Scholar
Danckwerts, P V (1951) Significance of liquid film coefficients in gas absorption. Ind Engng Chem 43:1460–1467.CrossRefGoogle Scholar
Deacon, E L, (1977) Gas transfer to and across the air-water interface. Tellus 29:363–374.CrossRefGoogle Scholar
Erickson, D J, ΙII (1992) A stability-dependent model for air-sea gas exchange. Submitted to J Geophys ResGoogle Scholar
Etcheto J, Merlivat L (1988) Satellite determination of the carbon dioxide exchange coefficient at the ocean-atmosphere interface: a first step. J Geophys Res 93:15669–15678.CrossRefGoogle Scholar
Heimann, M, Monfray P (1990) Spatial and temporal variation of the gas exchange coefficient for CO2, 1: data analysis and global validation Internal report No 31, Max Planck Institut fur MeteorolgieGoogle Scholar
Jahne, B, Huber W, Dutzi A, Wais T, Ilmberger J (1984) Wind/wave experiment on the Schmidt number and wave-field dependence of air/water gas exchange In: Brutsaert W, Jirka GH (eds) Gas transfer at water surfaces. Reidel, Dordrecht p 303.Google Scholar
Jahne, B, Munnich KO, Bosinger R, Dutzi A, Huber W, Libner P (1987) On parameters influencing air-water gas exchange. J Geophys Res 92:1937–1949.CrossRefGoogle Scholar
Ledwell J R, (1982) “Gas transfer at the air-water interface”, PhD thesis, Harvard University.Google Scholar
Liss, PS, Merlivat L (1986) Air-sea gas exchange rates: introduction and synthesis. In: Buat-Menard P (ed) The role of air-sea exchange in geochemical cycling. Reidel, Dordrecht, p 113.Google Scholar
Liss, PS (1983) Gas transfer: experiments and geochemical implications. In: Liss PS, Slinn WGN (eds) Air-sea gas exchange of gases and particles. Reidel, Dordrecht, p 241.Google Scholar
Memery L, Merlivat L (1985) Modelling of gas flux through bubbles at the air-water interface. Tellus 37B:272–285.CrossRefGoogle Scholar
Merlivat L, Memery L (1983) Gas exchange across an air-water interface: experimental results and modelling of bubble contribution to transfer J Geophys Res 88:707–724.CrossRefGoogle Scholar
Monahan EC, Muirchaertaich IO (1980) Optimal power law description of oceanic whitecap coverage dependence on wind speed. J Phys Oceanogr 10:2094–2099.CrossRefGoogle Scholar
Monahan EC, Spillane MC (1984) The role of oceanic whitecaps. In: air-sea gas exchange In: Brutsaert W, Jirka GH (eds) Gas transfer at water surfaces, Reidel, Dordrecht. p495.Google Scholar
Roether W, Kromer B (1984) Optimal application of the radon deficit method to obtain air-sea gas exchange rates. In: Gas transfer at water surfaces (W Brutsaert and G H Jirka, eds) Reidel, Dordrecht, pp 447–457.Google Scholar
Sarmiento JL, Orr JC, Siegenthaler U (1991) A perturbation model model of CO2 uptake in an Ocean General Circulation Model. J Geophys. Res (in press)Google Scholar
Thorpe SA (1982) The role of bubbles produced by breaking waves in supersaturating the near-surface mixed layer with oxygen. Ann. Geophys 2:53–56Google Scholar
Trenburth KE, Large WG, Olson JG (1989) The effective drag coefficient for evaluating wind stress over the oceans. J Climate 2:1507–1516CrossRefGoogle Scholar
Upstill-Goddard RC, Watson AJ, Liss PS, Liddicoat MI (1990) Gas transfer velocities in lakes measured with SF6. Tellus 42B:364–377Google Scholar
Wanninkhof R, Ledwell JR, Broecker WS (1985) Gas exchange-wind speed relation measured with sulphur hexafluoride on a lake. Science 227:1224–1226CrossRefGoogle Scholar
Wanninkhof R, Ledwell JR, Broecker WS, Hamilton M (1987) Gas exchange on Mono Lake and Crowley lake, California. J Geophys Res 92:14567–14580CrossRefGoogle Scholar
Wanninkhof R, Bliven LF (1989) Relationship between gas exchange, wind speed and radar backscatter. EOS Trans Am Geophys Union 70:297Google Scholar
Wanninkhof, R, (1992) Relationship between gas exchange and wind speed over the ocean Submitted to J Geophys ResGoogle Scholar
Watson AJ, Upstill-Goddard RC, Liss PS (1991) Air-sea gas exchange in rough and stormy seas measured by a dual-tracer technique. Nature 349:145–147.CrossRefGoogle Scholar
Woolf DK, Thorpe SA (1991) Bubbles and the air-sea exchange of gas in near-saturation conditions. J Mar Res 49:435–466.CrossRefGoogle Scholar