Machine Learning for Detection and Forecasting of Aerosol Events Using Long-Term Aerosol Measurements from Coastal, Rural and Urban Environments

Información del proyecto

NPF-ML

Identificador del acuerdo de subvención: 101284681

DOI

10.3030/101284681

Fecha de la firma de la CE 6 Agosto 2026

Fecha de inicio 30 Noviembre 2026

Fecha de finalización 29 Noviembre 2028

Financiado con arreglo a

• Marie Skłodowska-Curie Actions (MSCA)

Coste total

Sin datos

Aportación de la UE

€ 196 976,16

Coordinado por

VALSTYBINIS MOKSLINIU TYRIMU INSTITUTAS FIZINIU IR TECHNOLOGIJOS MOKSLU CENTRAS
Lithuania

Objetivo

Atmospheric ultrafine particles strongly impact climate forcing and human health, yet routine monitoring rarely captures their rapid formation. This project will develop a physics-informed AI framework to detect and forecast aerosol events, focusing on new particle formation (NPF) and short-term pollution episodes in varied environments. It leverages decade-long, high-resolution datasets from Lithuania’s Center for Physical Sciences and Technology (FTMC), including coastal Preila, urban Vilnius, and rural Rugšteliškis sites. Measurements from Scanning Mobility Particle Sizers, Aerodynamic Particle Sizers, and Aethalometers provide size-resolved particle distributions and black carbon data, enabling derivation of microphysical variables like condensation sink and growth rates. Methodology includes (i) applying convolutional neural networks (CNNs) to multi-year aerosol size-distribution ‘images’ for automated event detection; (ii) developing long short-term memory (LSTM) and hybrid CNN-LSTM models for 1–3 day forecasting; (iii) using transfer learning to generalize across sites; (iv) integrating anomaly detection for rare pollution episodes; and (v) employing probabilistic methods to quantify uncertainty and improve interpretability. The novelty is a transferable multi-site deep learning framework merging data-driven methods with aerosol physics, relying on ground-based instruments already used across Europe. Expected outcomes are open-source models, annotated datasets, and a prototype early-warning tool for agencies. This will enhance understanding of coastal vs urban aerosol dynamics, support climate and air quality policies, and boost public health preparedness. The fellowship will also advance the researcher’s career through machine learning training and integration into FTMC’s European networks like Aerosol, Clouds and Trace Gases Research Infrastructure (ACTRIS).

Ámbito científico (EuroSciVoc)

• ciencias naturales › informática y ciencias de la información › inteligencia artificial › aprendizaje automático › aprendizaje por transferencia
• ingeniería y tecnología › ingeniería ambiental › ingeniería de la contaminación atmosférica
• ciencias naturales › informática y ciencias de la información › inteligencia artificial › inteligencia computacional

Palabras clave

• Atmospheric Aerosols
• New Particle Formation
• Air Quality and Pollution Episodes
• Deep Learning
• Environmental Monitoring
• Anomaly Detection and Forecasting

Programa(s)

• HORIZON.1.2 - Marie Skłodowska-Curie Actions (MSCA) PROGRAMA PRINCIPAL
  

Tema(s)

• HORIZON-MSCA-2025-PF-01-01 - MSCA Postdoctoral Fellowships 2025
  

Régimen de financiación

HORIZON-TMA-MSCA-PF-EF - HORIZON TMA MSCA Postdoctoral Fellowships - European Fellowships

Convocatoria de propuestas

HORIZON-MSCA-2025-PF

Coordinador

VALSTYBINIS MOKSLINIU TYRIMU INSTITUTAS FIZINIU IR TECHNOLOGIJOS MOKSLU CENTRAS

• Aportación neta de la UEn: € 196 976,16
• Dirección: Savanoriu 231 ; 02300 Vilnius ; Lituania
• Región: Lietuva › Sostinės regionas › Vilniaus apskritis
• Tipo de actividad: Research Organisations
• Coste total: Sin datos