Differentiation of primary and secondary marine organic aerosol with machine learning

Abstract

Abstract. Marine primary organic aerosols (POAs) are important components of the marine climate system, regulating solar radiation budget and cloud dynamics. Despite their importance, there is a lack of extensive long-term observations of POA properties, introducing great uncertainty in their parameterization in models. This lack of information originates from the complexity of POA chemical composition, very few long-term high-resolution measurements of clean marine air, and the difficulty in performing source apportionment techniques over a long-term period. In this study, we utilize a comprehensive high-resolution time-of-flight aerosol mass spectrometer (HR-ToF-AMS) dataset spanning 1 decade (2009–2018) and introduce a machine learning (ML) approach to distinguish between marine POA and marine secondary organic aerosol (SOA). Results indicate that marine POA concentrations peak during summer months and reach their lowest levels in winter. On average, marine POA constitutes 51 % (ranging from 21 % to 76 %) of the marine organic aerosol (OA) annually and up to 63 % (48 % to 75 %) in summer. With the differentiated POA and SOA, we found diverse impacts of POA and SOA on aerosol hygroscopicity and mixing state. An increase in POA reduces the hygroscopicity and leads to an external state of mixing, while an increase in SOA sustains the relatively high hygroscopicity and leads to internal mixing. This study provides an observational dataset for marine POA and SOA and their diverse impacts on aerosol hygroscopicity, emphasizing a better appreciation of marine POA and SOA to improve the climate projections.