Structure, variability and persistence of the submicrometre marine aerosol

Submicrometre dry number size distributions from four marine and one continental aerosol experiment were evaluated jointly in the present study. In the marine experiments only data with back trajectories of at least 120 h without landcontact were used to minimize continental contamination. Log-normal functions were fitted to the size distributions. Basic statistics of the marine aerosol indicate a closed character of the size distribution at the lower size limit as opposed to an open character for corresponding continental data. Together with the infrequent occurrences of marine particles below 20 nm this finding supports hypotheses and model results suggesting lowprobabilities of homogeneous nucleation in the marine boundary layer. The variability of submicrometre marine number concentrations was parametrized with a bimodal log-normal function that quantifies the probability of finding different number concentrations about a given median value. Together with a four-modal log-normal approximation of the submicrometre marine size distribution itself, this model allows a statistical representation of the marine aerosol that facilitates comparison of experiments and validation of aerosol models. Autocorrelation at the one fixed marine site with a minimum of interruptions in times series revealed a strong size dependency of persistence in particle number concentration with the shortest persistence at the smallest sizes. Interestingly, in the marine aerosol (at Cape Grim) persistence exhibits a size dependency that largely matches the modes in d_g0, i.e. near the most frequent geometric mean diameters number concentrations aremost persistent. Over the continent, persistence of particle numbers is strongly constrained by diurnal meteorological processes and aerosol dynamics. Thus, no strong modal structure appears in the size-dependent persistence at Melpitz. As with the aerosol variability, marine aerosol processes in models of aerosol dynamics can be tested with these findings.