About me

Machine learning scientist working on forecasting atmospheric composition.

Now

I am a machine learning scientist at ECMWF, the European Centre for Medium-Range Weather Forecasts, working on atmospheric composition for the Copernicus Atmosphere Monitoring Service (CAMS). I work on AIFS-COMPO, a data-driven system for forecasting aerosols, reactive gases and air quality, which is developed with many colleagues at ECMWF. My work covers training data, model design and evaluation against observations, within the open-source Anemoi framework.

Before

  • Mila, MontréalPostdoc with David Rolnick on generative models for downscaling weather and climate data.
  • PhDMachine learning and computer science at Fraunhofer ITWM and the University of Kaiserslautern, together with the University of Oxford. I used deep learning to speed up climate models, supervised by Janis Keuper, Nicolas Gauger and Philip Stier.
  • InternshipsNASA/ESA Frontier Development Lab, Mila, and the Allen Institute for AI.
  • IndustryDeveloper and data scientist in the automotive industry.
  • TübingenMaster's degree in mathematics, specialising in numerical analysis with Christian Lubich.

Research interests

Machine learning for environmental, weather and climate modelling, in particular physical consistency, generalisation under distribution shift, and the representation of extreme events.

Selected papers

AIFS-COMPO: A Global Data-Driven Atmospheric Composition Forecasting System

Paula Harder, Johannes Flemming, Mihai Alexe, Gert Mertes, Baudouin Raoult, Matthew Chantry

arXiv preprint · 2026 arXiv ECMWF news EGU 2026

Reconstructing Carbon Monoxide Reanalysis with Machine Learning

Paula Harder, Johannes Flemming

arXiv preprint · 2026 arXiv EGU 2026

Benchmarking the geographic generalization of deep learning models for precipitation downscaling

Paula Harder, Luca Schmidt, Francis Pelletier, Nicole Ludwig, Matthew Chantry, Christian Lessig, Alex Hernandez-Garcia, David Rolnick

Scientific Reports · 2026 Sci. Rep. arXiv (RainShift) code dataset

On Global Applicability and Location Transferability of Generative Deep Learning Models for Precipitation Downscaling

Paula Harder, Christian Lessig, Matthew Chantry, Francis Pelletier, David Rolnick

arXiv preprint · 2025 arXiv EGU 2025

Hard-constrained deep learning for climate downscaling

Paula Harder, Alex Hernandez-Garcia, Venkatesh Ramesh, Qidong Yang, Prasanna Sattigeri, Daniela Szwarcman, Campbell Watson, David Rolnick

Journal of Machine Learning Research (JMLR) · 2023 JMLR arXiv code

Physics-informed learning of aerosol microphysics

Paula Harder, Duncan Watson-Parris, Philip Stier, Dominik Strassel, Nicolas R. Gauger, Janis Keuper

Environmental Data Science · 2022 EDS code arXiv

All publications