Detailed programme of the MOOC

Tier 1 - ML in Weather & Climate

In this first tier of the "Machine Learning in Weather & Climate" MOOC, we introduce key concepts in machine learning, and numerical weather and climate prediction. This includes the main topics, from the processing of observations, to data assimilation, forecasting and post-processing. In each element the relevance of machine learning is described, to understand the added value and limitations of machine learning in weather and climate.

Introduction to ML in Weather & Climate

Launch date: Monday, 9 January 2023
Duration: 6h e-learning and 3 webinars
This module will introduce Numerical Weather Prediction (NWP), Machine Learning (ML) and climate challenges. You will discover the various types of ML, why they are used in NWP, and the challenges in applying them. You will hear directly from the experts and have a chance to interact with them in a number of live events, including a dedicated session with industry representatives.
Core lessons in this introductory module cover the topics of:

  • Introduction to numerical weather and climate predictions, machine learning, big data and high performance computing
  • Expert opinions on Machine Learning, how it changes the state of the art and what is needed to make the most of it
  • Types of machine learning, and their applications
  • Challenges for the use of machine learning in weather and climate predictions
  • Applying AI responsibly for environmental sciences
Experts: F. Rabier, S. Burgess, P. Dueben, J. Dramsch, M. Chantry (ECMWF), F. Ubertini (IFAB/CINECA), E. Barnes (University of Colorado), J. Kristiansen (MetNorway), M. Kuglitsch (Fraunhofer Society), H. Hoos (RWTH Aachen University), A. McGovern (University of Oklahoma), M. Reichstein (MPI-Jena), G. Camp-Valls (University of Valencia), T. McCaie (UK Met Office), L. de Cruz (Met Belgium), S. Materia (Barcelona Supercomputing Center), R. Schneider (ESA), T. Paccagnella (ARPAE), H. Evers-King (EUMETSAT), A.Volpe (BIP), P. Gambetti (CRIF), F. della Casa (Leithà), G. Jauvion (AccuWeather) and M. Abel (Ambrosys)


Launch date: Monday, 16 January 2023
Duration: 1h e-learning
Observations of our Earth are our way to measure and understand the current state of weather and climate. An accurate understanding and processing of this data enables us to make the best weather and climate predictions possible. However, obtaining, analysing and processing this data is not always straight-forward.
Core lessons in the observation module cover the topics of:

  • Use of observations in NWP and data assimilation
  • Data analysis in Numerical Weather Prediction (NWP)
  • Challenges in observation datasets
  • Machine Learning for data processing and observation operators
  • Machine Learning for the processing of satellite data
Experts: A. Geer, P. de Rosnay (ECMWF), D. Tuia (Aachen University) and M. Higgins (EUMETSAT)

Forecast Model

Launch date: Monday, 23 January 2023
Duration: 3h30 e-learning and 1 webinar
The forecast model lies at the core of numerical weather and climate prediction. This model is run on powerful supercomputers to calculate how the weather and climate change from the current measurements worldwide. However, these models come in many shapes and forms from now-casting predictions for the next minutes to climate models in the next 30 years.
The lessons in the forecast model module cover the topics of:

  • Development of modern forecast models
  • Forecasting assisted with machine learning from emulation to hybrid modelling
  • Measuring the quality of machine learning forecasts with the WeatherBench benchmark
  • Looking ahead to Digital Twins of the Earth’s weather and climate
  • Explore how machine learning will be used in weather and climate forecasting in 10 years from now?
Experts: P. Dueben, M. Chantry, M. Clare, P. Bauer (ECMWF), K. Kashinath (NVIDIA), S. Ravuri (DeepMind) and R. Furner (BAS)

Data Assimilation

Launch date: Monday, 30 January 2023
Duration: 30mn e-learning
Data assimilation is the art and science of combining information from weather and climate models and observations to generate the best possible representation of the state of the Earth system. Data assimilation requires estimates of uncertainties and errors for both observations and the model state and the mathematical tools which are used are often complex.
Core lessons in the data assimilation module cover the topics of:

  • Data assimilation in numerical weather prediction (NWP)
  • Treatment of uncertainty from the model and observations
  • Modelling adjustments using Kalman filters and the so-called 4D-Var algorithm
  • Similarities between data assimilation and machine learning
Expert: M. Bonavita (ECMWF)


Launch date: Monday, 6 February 2023
Duration: 1h e-learning
George Box said it well: "All models are wrong, but some are useful". As good as our weather and climate predictions are, they could always be better. For this reason, we often analyse historical model forecasts to understand what they got wrong, then we can apply those corrections to the latest forecasts to improve their accuracy.
Core lessons in the post-processing module cover these topics:

  • Core principles of weather and climate post-processing
  • Post-processing with machine learning
  • The World Meteorological Organisation (WMO) seasonal to subseasonal (S2S) weather prediction challenge
  • From the model to the end-user – how do forecast simulations on a supercomputer end up as weather forecasts on your mobile?
  • Apply simple post-processing for weather output data over your hometown
Experts: F. Vitart, B. Hemingway (ECMWF) and S. Lerch (KIT)


Launch date: Monday, 13 February 2023
Duration: 1h30 e-learning
The days of pen and paper are gone. Modern weather and climate predictions are calculated on the largest high-performance supercomputers in the world. Software and hardware play a central role in the daily work of operational meteorologists. And yes, before anyone asks, we will talk about cloud computing.
Core lessons in the software & hardware module cover these topics:

  • Hardware infrastructure from CPU to GPUs to quantum computing
  • Machine learning in software and hardware development
  • Introduction to the European Weather Cloud
  • The new supercomputer at ECMWF
  • Supercomputer and machine learning – a golden combination
  • Custom hardware solutions for modern compute requirements
Experts: C. Kitchen, P. Dueben, V. Baousis, I. Hadade, M. Dell'Acqua (ECMWF) and M. Brorsson (University of Luxembourg)

Download the programme

See Tier 2 >