Cloud condensation nuclei in polluted air and biomass burning smoke near the mega-city Guangzhou, China – Part 2: Size-resolved aerosol chemical composition, diurnal cycles, and externally mixed weakly CCN-active soot particles

摘要

Size-resolved chemical composition, mixing state, and cloud condensation nucleus (CCN) activity of aerosolparticles in polluted mega-city air and biomass burning smoke were measuredduring the PRIDE-PRD2006 campaign near Guangzhou, China, using an aerosolmass spectrometer (AMS), a volatility tandem differential mobility analyzer(VTDMA), and a continuous-flow CCN counter (DMT-CCNC).The size-dependence and temporal variations of the effective averagehygroscopicity parameter for CCN-active particles (κ_a) could beparameterized as a function of organic and inorganic mass fractions (f_org, f_inorg)determined by the AMS: κ_a,p=κ_org·f_org +κ_inorg·f_inorg. The characteristic κ values oforganic and inorganic components were similar to those observed in othercontinental regions of the world: κ_org≈0.1 andκ_inorg≈0.6. The campaign average κ_avalues increased with particle size from ~0.25 at ~50 nm to~0.4 at ~200 nm, while f_org decreased with particle size.At ~50 nm, f_org was on average 60% and increased to almost100% during a biomass burning event.The VTDMA results and complementary aerosol optical data suggest that thelarge fractions of CCN-inactive particles observed at low supersaturations(up to 60% at S≤0.27%) were externally mixed weakly CCN-active soot particles with lowvolatility (diameter reduction <5% at 300 °C) and effectivehygroscopicity parameters around κ_LV≈0.01. A proxy forthe effective average hygroscopicity of the total ensemble of CCN-active particlesincluding weakly CCN-active particles (κ_t) could be parameterized as a functionof κ_a,p and the number fraction of low volatility particlesdetermined by VTDMA (φ_LV): κ_t,p=κ_a,p−φ_LV·(κ_a,p−κ_LV).Based on κ values derived from AMS and VTDMA data, the observed CCNnumber concentrations (N_CCN,S≈102–104 cm?3 atS = 0.068–0.47%) could be efficiently predicted from the measured particlenumber size distribution. The mean relative deviations between observed andpredicted CCN concentrations were ~10% when using κ_t,p,and they increased to ~20% when using only κ_a,p. The meanrelative deviations were not higher (~20%) when using an approximatecontinental average value of κ≈0.3, although the constantκ value cannot account for the observed temporal variations inparticle composition and mixing state (diurnal cycles and biomass burningevents).Overall, the results confirm that on a global and climate modeling scalean average value of κ≈0.3 can be usedfor approximate predictions of CCN number concentrations in continentalboundary layer air when aerosol size distribution data are available withoutinformation about chemical composition. Bulk or size-resolved data on aerosolchemical composition enable improved CCN predictions resolving regional and temporalvariations, but the composition data need to be highly accurate andcomplemented by information about particle mixing state to achieve highprecision (relative deviations <20%).