Gradient flux measurements of sea–air DMS transfer during the Surface Ocean Aerosol Production (SOAP) experiment

Direct measurements of marine dimethylsulfide (DMS) fluxes are sparse, particularly in the Southern Ocean. The Surface Ocean Aerosol Production (SOAP) voyage in February–March 2012 examined the distribution and flux of DMS in a biologically active frontal system in the southwest Pacific Ocean. Three distinct phytoplankton blooms were studied with oceanic DMS concentrations as high as 25 nmol L−1. Measurements of DMS fluxes were made using two independent methods: the eddy covariance (EC) technique using atmospheric pressure chemical ionization–mass spectrometry (API-CIMS) and the gradient flux (GF) technique from an autonomous catamaran platform. Catamaran flux measurements are relatively unaffected by airflow distortion and are made close to the water surface, where gas gradients are largest. Flux measurements were complemented by near-surface hydrographic measurements to elucidate physical factors influencing DMS emission. Individual DMS fluxes derived by EC showed significant scatter and, at times, consistent departures from the Coupled Ocean–Atmosphere Response Experiment gas transfer algorithm (COAREG). A direct comparison between the two flux methods was carried out to separate instrumental effects from environmental effects and showed good agreement with a regression slope of 0.96 (r2 = 0.89). A period of abnormal downward atmospheric heat flux enhanced near-surface ocean stratification and reduced turbulent exchange, during which GF and EC transfer velocities showed good agreement but modelled COAREG values were significantly higher. The transfer velocity derived from near-surface ocean turbulence measurements on a spar buoy compared well with the COAREG model in general but showed less variation. This first direct comparison between EC and GF fluxes of DMS provides confidence in compilation of flux estimates from both techniques, as well as in the stable periods when the observations are not well predicted by the COAREG model.