About - The MLPP in a nutshell
The Machine Learning Pilot Project (MLPP) is a collaborative initiative led by Met Norway and MeteoSwiss, involving 15 partners from 14 countries, with the goal of integrating machine learning (ML) into numerical weather prediction (NWP). The project is structured into four key areas: data-driven forecasting, ensemble forecasting, data assimilation and infrastructure/MLOps. It was kicked off in June 2024 and will run until July 2027.
The project aims to advance data-driven weather models by leveraging different datasets such as re-analysis, operational analysis and forecasts and observations. It also seeks to develop ML-based ensemble forecasting techniques to improve the reliability of probabilistic predictions. Another core focus is enhancing data assimilation through ML to improve initial conditions in NWP. Additionally, the project is working on building infrastructure to support ML workflows, addressing hardware compatibility, software gaps and model deployment.
Project structure:
(WP0 Coordination)
WP1 Data driven modelling
WP2 Ensemble forecasting
WP3 Data assimilation
WP4 MLOps and infrastructure (cross-cutting)
WP5 Training and support (cross-cutting)
Do you want to join us? National Meteorological Services of ECMWF Member States can join the project as a partner - for more information, please contact the project managers Jorn Kristiansen (jornk(at)met.no) or Katrin Ehlert (katrin.ehlert(at)meteoswiss.ch).
Add Comment