Investigation of global aerosol regime developments from pre-industrial times to the future

Abstract

The temporal development of the global distribution of aerosol regimes, i.e. areas of specific aerosol characteristics, is investigated based on global aerosol simulations performed for six time slices, including pre-industrial (1750, 1850) and present-day conditions (2015), as well as future (2050) following the Shared Socioeconomic Pathways SSP3-7.0, SSP2-4.5 and SSP1-1.9. The aerosol characteristics are evaluated in terms of aerosol regimes by combining the information about multiple aerosol properties within a cluster analysis. The present-day case serves as a reference where aerosol regimes are identified via the unsupervised machine learning technique K-means. In addition, a supervised machine learning technique (Random Forest) is applied on the basis of the aerosol regime classification criteria gained from the reference case. This allows for a consistent identification of comparable aerosol regimes across the different time slices. With this approach we analyze, for instance, which present-day aerosol regimes could have been present in the past and how they might develop in the future. These analyses are conducted for three atmospheric altitude layers with a special focus on the lower troposphere. After conducting a primary classification, we also implement a secondary sub-classification to assess the fine structure of the identified primary regimes. The results are further complemented by an in-depth analysis of the emissions of primary aerosol species and aerosol precursor gases for the different time slices in selected regions, including information about the respective contributions from different emission sectors. Details will be presented on a poster.