KCDS Virtual Open House 2025, Fall edition: Project 01 - Estimation of short-term extremes in air pollution and their drivers

Estimation of short-term extremes in air pollution and their drivers
Air pollution is a major problem in cities around the world. Extremely polluted outdoor air is a particular health hazard, even short-term exposure to high pollution levels causes about 1 million premature deaths annually. An early warning would help mitigate these situations but requires high-accuracy prediction. Numerical weather models are able to forecast meteorological conditions with good accuracy. However, the complex interaction between the various pollution sources and factors such as wind (for transport), precipitation (removal of pollutants), and vegetation (source and sink of particles depending on season) poses problems for the prediction of extreme air pollution with weather models.
The aim of this project is to develop a new data-driven approach to accurately predict extremes in urban air pollution from observation data, and to quantify the contributions of the various factors driving these extremes. Therefore, methods from extreme value statistics will be implemented in a machine-learning framework to estimate and predict air pollution extremes. As a starting point, we use Paris as a baseline city, where ML methods were successfully implemented to reproduce specified PM1 concentrations. ... mehrThis model architecture will be adapted and retrained on multi-city data from global air pollution hotspots, enabling generalization beyond the urban area of Paris.
The project lies at the intersection of innovation in mathematics and meteorology. By combining extreme value theory with machine learning, it aims to advance the current state of the art in predicting air pollution extremes.
The doctoral degree will be awarded in either Mathematics or Meteorology, depending on the candidate’s academic background.
Special requirements for applicants:
• Outstanding Master's degree in Mathematics, Statistics, Meteorology or related field.
• Demonstrated interest in the application of mathematical methods to meteorological or environmental phenomena.
• Solid mathematical background.
• High motivation to apply supervised learning methods.
• Excellent command of English, both spoken and written.