Internship title: Machine Learning for Africa: Deploying Machine Learning Models for Weather Prediction in Africa

EUMETSAT has a close cooperation with Africa, we have more than 30 years of cooperation support weather sevcies to use EUMETSAT data in forecasting operations to support the protection of life and property and economic development. Currently EUMETSAT is responsible to supporting the implementation of the AMSAF- nowcasting in cooperation with the African Regional Specialised Meteorological Centres with responsibility for nowcasting. At the core of te AMSAF-nowcasting is the use of the Nowcasting SAF software to produce products for the regional centres. This project is funded by the European Union and is aligned with EUMETSAT objectives.

There are a number of machine learning models such as Rainfall for Africa (Chalmers University) and NCAST (University of Leeds) that are driven by EUMETSAT data and targeted at African users. In addition, there are some centres working on global models such as MAZU in China.

EUMETSAT is interested in understanding the maturity and applicability of these models to support the African forecast centres. It may be anticipated that as these models mature, they will provide faster methods to produce nowcasting and associated products; for example, tropical cyclone monitoring, dust, fire or atmospheric composition.

At this time there is no toolkit available to support the validation, benchmarking and comparison of these, and future, models so that the quality can be assed prior to potential transfer to operations This would be a core part of the internship: to establish workflows in the European Weather Cloud to deploy such models, to run inference with EUMETSAT data and enable such a comparison. This would then enable faster uptake of appropriate models for the African users, and put in allow for the extension of the workflows to cover validation for Member State application and application in other thematic areas.

The intern will perform the following tasks:

• Perform research in existing Machine Learning Models for Weather Prediction for Africa. Make an overview of models for weather predictions for Africa and their input data. Make an assessment on which models can run with EUMETSAT data.
• Deploy selected models on the European Weather Cloud.

• Run inference of the models with data from EUMETSAT (FCI, GEORing). Optionally, extend the inference approach to run on near-real-time (NRT) data from EUMETSAT.
• Create a small application or Jupyter Notebook for visualizing the model output.
• Present the outcomes and results within the organization and beyond.

• Have the ability to work effectively in English;
• Be computer literate;
• Be intrinsically motivated and curious about the internship subject;
• Be able to work independently and collaboratively;
• Have the ability to take the initiative in researching ideas;
• Have the ability to collect, collate, conceptualize and present information clearly.

Additional advantageous requirements:

• Programming experience, preferably in Python;
• Experience with Machine Learning (ML);
• Knowledge in EO data or Earth Sciences;
• Relevant project or work experience in a technical field through class projects, clubs or internships;
• Experience writing code and contributing to open source projects.