Experimental studies of the parameters controlling bubble-induced air-sea gas exchange

<p>An experimental study for the investigation of bubble-mediated gas exchange in a dedicated bubble tank is presented. Bubbles were generated by a water jet with adjustable kinetic energy, which entrained a controllable gas volume flux into the water tank. Bubble size distributions in a radius range from 10 μm to 5000 μm and velocity distributions in a range from 70 μm to 3000 μm were measured at 60 positions in the tank by a telecentric optical setup.</p> <p>Trace gas concentrations were measured using a quadrupole mass spectrometer with a silicone membrane inlet in the water phase of a gas tight bubble tank. Fourteen trace gases with a wide range of solubilities and diffusivities - SF6, Ne, N2 ,HD, D2, O2, Kr, Pentafluoroethane, Xe, N2O, C2H2 , CH3Cl, Benzene and DMS - were used to investigate the dependency on these two physico-chemical parameters. Evasion experiments with a variety of conditions were conducted, including salt water (1.75 % NaCl), the addition of the soluble surfactant Triton X- 100, n-butanol and glycerol. All used additives tend to increase the fraction of small bubbles generated and therefore increase the transfer velocity at constant gas volume flux and energy<br> input. The addition of salt had the strongest effect, leading to an enhancement of the transfer velocity up to a factor of ~4. The increase of the transfer velocity adding Triton X-100 was found<br> to be lower than 5 %. The transition between diffusivity controlled and solubility controlled transfer occurs at solubilities ranging from 0.05-0.44. The transition is shifted to lower solubilities by a factor of ~3 by the addition of salt. Existing models for bubble mediated gas transfer are tested. Simple power law dependencies turn out to be incapable to describe the transfer for the whole range of solubilities and diffusivities. An extension of the parameterization proposed by Woolf et al. (2007) requiring 4 parameters fits the data best. A simple model using only 2 parameters is proposed. Its performance is almost as good as the extended Woolf model.