Secondary organic aerosol formation from ambient air in an oxidation flow reactor in central Amazonia

Secondary organic aerosol (SOA) formation from ambient air was studied using an oxidation flow reactor (OFR) coupled to an aerosol mass spectrometer (AMS) during both the wet and dry seasons at the Observations and Modeling of the Green Ocean Amazon (GoAmazon2014/5) field campaign. Measurements were made at two sites downwind of the city of Manaus, Brazil. Ambient air was oxidized in the OFR using variable concentrations of either OH or O-3, over ranges from hours to days (O-3) or weeks (OH) of equivalent atmospheric aging. The amount of SOA formed in the OFR ranged from 0 to as much as 10 mu g m(-3), depending on the amount of SOA precursor gases in ambient air. Typically, more SOA was formed during nighttime than daytime, and more from OH than from O-3 oxidation. SOA yields of individual organic precursors under OFR conditions were measured by standard addition into ambient air and were confirmed to be consistent with published environmental chamber-derived SOA yields. Positive matrix factorization of organic aerosol (OA) after OH oxidation showed formation of typical oxidized OA factors and a loss of primary OA factors as OH aging increased. After OH oxidation in the OFR, the hygroscopicity of the OA increased with increasing elemental O : C up to O : C similar to 1.0, and then decreased as O : C increased further. Possible reasons for this decrease are discussed. The measured SOA formation was compared to the amount predicted from the concentrations of measured ambient SOA precursors and their SOA yields. While measured ambient precursors were sufficient to explain the amount of SOA formed from O-3, they could only explain 10-50% of the SOA formed from OH. This is consistent with previous OFR studies, which showed that typically unmeasured semivolatile and intermediate volatility gases (that tend to lack C=C bonds) are present in ambient air and can explain such additional SOA formation. To investigate the sources of the unmeasured SOA-forming gases during this campaign, multilinear regression analysis was performed between measured SOA formation and the concentration of gas-phase tracers representing different precursor sources. The majority of SOA-forming gases present during both seasons were of biogenic origin. Urban sources also contributed substantially in both seasons, while biomass burning sources were more important during the dry season. This study enables a better understanding of SOA formation in environments with diverse emission sources.