Conference Agenda
| Session | ||||
WG5: Aerosol transport and data science
| ||||
| Presentations | ||||
10:15am - 10:30am
WE1-3: 1 A coupled Lagrangian-Equilibrium approach for the simulation of volcanic aerosol plumes 1Department of Mechanical Engineering, Imperial College London, London, UK; 2Department of Aeronautics, Imperial College London, London, UK; 3Met Office, Exeter, UK In this work, we propose a formulation for two-way coupled large eddy simulations of particle laden dilute flows, with the aim of simulating volcanic aerosol plumes.The method comprises a stochastic Lagrangian approach coupled with the equilibrium model for low-inertia particles (Stokes number less than 1). This allows for capturing events in regions where the applicability of the equilibrium model is ambiguous and retain the benefits of using a Eulerian approach, which include facilitating the modelling of microphysical kinetic processes of aerosol interaction such as aggregation, while keeping the computational cost and memory requirements low.
10:30am - 10:45am
WE1-3: 2 Towards atmospheric compound identification in chemical ionization mass spectrometry with machine learning 1Department of Chemistry, University of Helsinki, 00560 Helsinki, Finland; 2Department of Applied Physics, Aalto University, Espoo, Finland; 3Aerosol Physics Laboratory, Physics Unit, Tampere University, 33720 Tampere, Finland; 4Karsa Ltd., A. I. Virtasen aukio 1, 00560 Helsinki, Finland; 5Physics Department, TUM School of Natural Sciences, Technical University of Munich, Garching, Germany; 6Atomistic Modelling Center, Munich Data Science Institute, Technical University of Munich, Garching, Germany; 7Munich Center for Machine Learning (MCML) Chemical ionization mass spectrometry (CIMS) is essential in atmospheric chemistry research but faces challenges in compound identification due to complex reagent ion-target compound interactions. Quantum chemical calculations can model these interactions, yet the vast configuration space and high costs hinder database creation, preventing a definitive compound identification workflow. This project explores a machine learning (ML) cost-efficient approach for CIMS compound identification. As a first step, the ML workflow developed can predict the detection and signal intensity of known compounds, and map the functional groups likely interacting with the reagent ion.
10:45am - 11:00am
WE1-3: 3 The Role of Hygroscopic Properties in Nitrate Formation and Its Impact on Haze in Seoul 1Graduate School of Public Health, Seoul National University; 2Institute of Health and Environment, Seoul National University Severe PM pollution episodes frequently impact the Seoul Metropolitan Area, with nitrate playing a dominant role. Nitrate formation occurs through complex atmospheric pathways, including gas-phase oxidation, heterogeneous uptake, and aqueous-phase processes. This study employs explainable machine learning to analyze drivers of nitrate formation in Seoul, using high-resolution aerosol mass spectrometry data. Results reveal that nitrate formation sharply increases at RH >65%, persisting even as RH declines, suggesting prolonged aqueous-phase processing. Additionally, temperature exhibits a strong nonlinear effect, with nitrate formation suppressed below freezing due to phase-state constraints. These findings highlight the role of hygroscopic properties in nitrate-driven haze.
11:00am - 11:15am
WE1-3: 4 Increasing Impact of Transported Dust to Europe in a Changing Climate Paul Scherrer Institute, Switzerland In this work, we present the most complete database of transported metals, along with a machine learning model designed to predict dust concentrations over Europe. The model is validated thoroughly with a combination of long measurement timeseries and ice core records. We find that dust concentrations increased over the period of 2012-2021, driven by more severe episodes, as a consequence of further desertification, which has significant implications both for regulatory purposes but also for public health.
| ||||