MOOC on machine learning in weather and climate

The Massive Open Online Course (MOOC) on Machine Learning in Weather and Climate was the first MOOC carried out by ECMWF. It aimed to train a wider community on the impact and use of machine learning in numerical weather and climate predictions.

Sixty experts from ECMWF and other organisations contributed to the course materials. Pictured here are some of the ECMWF contributors; left to right: Matthew Chantry, Mariana Clare, Jesper Dramsch, Peter Düben, Siham El Garroussi.

Sixty experts from ECMWF and other organisations contributed to the course materials. Pictured here are some of the ECMWF contributors; left to right: Matthew Chantry, Mariana Clare, Jesper Dramsch, Peter Düben, Siham El Garroussi.

9,026
Participants
159
Countries
60
Experts
40
Hours of content
47
Videos
15
Webinars
37
eLearning modules
1,147
Participants 1st webinar
1,235
Forum messages
6,021
Certificates
4.3/5
Satisfaction rate

The MOOC ran from 9 January to 30 April 2023 in partnership with the International Foundation on Big Data and Artificial Intelligence for Human Development (IFAB). It brought together experts throughout Member and Co-operating States and beyond to provide a shared vision across the communities of Earth system sciences, high-performance computing and machine learning.

The free online training course was accessible to a global audience. To maximise inclusivity, participants could follow one or more tiers and select optional modules within each tier according to their interests. To allow for self-paced learning, participants could complete at least four hours of study per week, for a total of at least 40 hours of training.

Tier 1, ‘Machine Learning in Weather and Climate’, was an introduction, aimed at anyone interested in the topic. Only a basic knowledge of weather and climate science, statistics and computing was assumed. The programme covered the use of machine learning across various themes, from the processing of observations to data assimilation, forecasting and post-processing.

9,026
Participants
159
Countries
60
Experts
40
Hours of content
47
Videos
15
Webinars
37
eLearning modules
1,147
Participants 1st webinar
1,235
Forum messages
6,021
Certificates
4.3/5
Satisfaction rate

Tier 2, ‘Concepts of Machine Learning’, delved into the technical aspects of machine learning. It was a lot more hands-on, including coding assignments using Jupyter notebooks in addition to eLearning modules covering the theory behind machine learning algorithms. This tier was more targeted to technical data users from academia or industry across different sectors.

The aim of tier 3, ‘Practical Machine Learning applications in weather and climate’, was to demonstrate how the techniques presented in tier 2 could be practically applied to real-world applications across the same topics introduced in tier 1. The end of the MOOC coincided with the application phase of ECMWF’s Code for Earth programme, and participants were encouraged to submit proposals for coding projects with ECMWF mentoring and the chance to win a cash stipend.

More than 6,000 certificates were issued for completion of one or more tiers. After the live run had ended, the MOOC materials remained freely accessible in ECMWF’s online eLearning resources. This was the first MOOC carried out by ECMWF and it established and cemented collaborations across the wider machine learning in weather and climate community. The interest generated by the initiative encouraged us to consider other topics for future MOOCs.