High-resolution modeling of organic aerosol and its components over Europe

Aerosol components have distinct health and climate impacts, hence considering them is essential to reduce uncertainty in estimating the health impacts attributed to aerosol. Organic aerosols (OAs) are one of the significant components of aerosol having a substantial share in total aerosol mass in Europe with significant seasonal and spatial variations. OAs are majorly emitted from fossil fuel combustion and generated through secondary aerosol formation too and based on the origin OA components can be categorized into biomass-burning organic aerosol (BBOA), hydrocarbon-like aerosol (HOA), and oxygenated organic aerosol (OOA).  OA measurements are limited, and measurements of its components are even more sparse due to the need for highly sophisticated instruments and advanced technical expertise. Additionally, OA and its components are not mandated for monitoring at all sites under government regulations, they are measured only at spatially sparse supersites. Chemical transport modeling provides spatially and temporally continuous OA and OA components with some vested uncertainty, but they are generally done at coarser resolution because of computational constraints. Whereas epidemiologist requires data at a finer resolution to link them with the local scale health data. Hence we are modeling total OA and OA components (BBOA, HOA, and OOA) at a fine resolution over Europe. Here, we use an integrated modeling approach combining a chemical transport model and machine learning algorithms. We simulated OA and its components using a comprehensive air quality model with extensions (CAMx) at around 10 km resolution over Europe for 10 years. In the subsequent step, a random forest (RF) model was trained using CAMx outputs, meteorological parameters, and land use variables as predictors to estimate observed aerosol component concentrations as the target to improve the predictions and enable downscaling of the outcome. Here we used an unparallel large OA and OA components observation inventory made with measurements from various research and operational groups across the world, consisting of 50,000 daily observations for OA and 15,000 daily observations for OA components from 140 and 40 locations respectively. This approach improves predictions with an increase in r2 to 0.43 from 0.31 for total OA and reduces the RMSE from  1.5, 0.8, 3.1 µg/m³ to 0.3, 0.2 and 0.45 µg/m³ for BBOA, HOA, and OOA respectively, while also enabling downscaling. This provided us with OA and OA components at a higher resolution of 200m across Europe for 10 years. The high-resolution modeled map illustrates the regional to local spatial distribution of OA and its components. Such high-resolution aerosol component modeling is instrumental for epidemiological studies when combined with local health datasets. Local-scale analysis with such datasets helps identify dominant aerosol components and their sources.