Modelling Aerosol Effects on Climate : From Organic Volatiles to Cloud Interactions and Radiative Forcing

Aerosols play a critical role in Earth’s climate by influencing cloud formation, atmospheric chemistry, and radiative balance. However, aerosol-cloud and aerosol-radiation interactions remain major sources of uncertainty, affecting both radiative forcing estimates and future climate projections. A significant fraction of atmospheric aerosols originates from secondary organic aerosol (SOA), which forms through the oxidation of volatile organic compounds (VOCs). Modelling SOA is challenging due to the multitude of oxidation products involved, which exhibit a wide range of volatilities that influence their gas-particle partitioning. Moreover, the volatilities of these oxidation products remain highly uncertain, making it difficult to model SOA formation and its impact on climate. In addition to SOA, primary aerosols from forest wood use contribute to atmospheric composition and climate, with their impact depending on how harvested wood is utilized. Key emissions from forest wood use, including black carbon (BC), organic carbon (OC), and sulphur dioxide (SO 2 ), contribute to radiative forcing and influence climate processes. Furthermore, the relationship between Cloud condensation nuclei (CCN) and cloud droplet number concentration (CDNC) is not straightforward, as it depends on multiple atmospheric factors beyond CCN alone. It is therefore essential to understand these interconnected processes, which are critical for improving the aerosol representation in climate models and reducing uncertainties in aerosol radiative forcing estimates.

This thesis investigates three key aspects of aerosol-climate interactions by integrating process-level modelling, global climate model simulations, satellite observations, and statistical analysis. First, we examine how the volatilities of SOA precursors influence particle growth to CCN sizes and radiative effects. This is done using the Volatility Basis Set (VBS) framework in both process-scale and global-scale models. The VBS approach represents organic compounds by grouping them into discrete volatility bins, improving the modelling of their role in the atmosphere. Second, we assess the impact of alternate forest biomass use in Finland on aerosol emissions and radiative forcing. Third, we investigate the limitations of traditional regression methods in estimating the correlation between CCN and CDNC and propose a machine-learning based approach to improve our understanding on aerosol-cloud interactions.

To address uncertainties in the volatility of VOC oxidation products, we performed global climate model simulations to evaluate how different volatility assumptions influence modelled aerosol effects on climate. Our results indicate that a ten-fold decrease in the assumed volatility increases SOA burden by 13 %, whereas, an increase in volatility reduces it by 9 %. This change exerts a moderate effect on CCN concentrations, with a decrease of 3 % for increased volatility and an increase of 2 % for reduced volatility relative to the original volatility. Comparing the binning of VBS to three and nine volatility classes reveals that the choice of the resolution of volatility bins substantially influences CCN estimates. Our study on radiative forcing from different wood use scenarios shows that while instantaneous radiative forcing (IRF ARI ) remains relatively small, aerosol-induced effective radiative forcing (ERF) can be significant due to changes in the allocation of additional wood harvest. The net ERF over Finland is +0.13 W m −2 for increased pulpwood harvest, whereas for increased energy biomass combustion in small-scale boilers, it reaches -0.90 W m −2 . Finally, we investigated the correlation between CCN and CDNC using traditional least square methods such as ordinary least squares in its two-dimensional form (OLS) and bivariate least squares (BLS). We found that while BLS accounts for uncertainties in both axes, it often produces unphysical slope estimates. Both methods are inherently bivariate and fail to consider additional confounding variables. These limitations underscore the need for a more comprehensive approach that accounts for key atmospheric factors to better quantify aerosol-cloud interactions. We propose the use of a more comprehensive approach, such as machine learning, to improve CCN-CDNC slope estimates.

Overall, this thesis underscores the persistent uncertainties in aerosol-cloud interactions and emphasizes the need for improved representation of organic aerosols in global-scale models. It highlights the importance of employing appropriate statistical methods and refining aerosol-cloud-radiation assessments to improve our understanding of their climatic impacts.

Keywords: Modelling, volatility, SOA, clouds, satellite, machine-learning, cloud parcel model, updraft velocity, forest biomass, emissions