Intern
    Data Science Chair

    Praktikum: Machine Learning for Time Series Analysis - AI Weather Quest

    Deep Learning for Climate Modelling

    Supervisors: Dr.-Ing. Anna Krause,
     

    Scope: 5 ECTS/10 ECTS

    Courses: 10-I=PRJAK / 10-I=PRAK, 10-I=PDS1, 10-I=PDS2, 10-I-PIS1, 10-I-PIS2,

    Kickoff: April 23rd 2025, 15:00-17:00, Seminar room 1, Emil-Fischer-Straße 50

    Enrollment: WueCampus Course with Password Pr_DLClim

    WueCampus Course: https://wuecampus.uni-wuerzburg.de/moodle/course/view.php?id=72983

    Content

    In this Master's practical course, the students take part in the European Center for Midrange Weather Forecasting's (ECMWF) AI Weather Quest. Students will develop and deploy Deep Learning (DL) Models for Subseasonal forecasting.

    Subseasonal forecasting refers to predicting weather two to four weeks in advance, and sits between weather forecasting and climate modeling. Currently, weather forecasting and climate modeling are ongoing topics in AI research, but the time in-between - seasonal and subseasonal forecasting is not being addressed by the AI research community. The AI Weather Quest challenge addresses thie modeling gap.

    The challenge runs in two phase:

    • the training phase until August 22nd: during this phase teams train their models and get one 13 week trial evaluation phase
    • the competition phase until August 2026: during this phase the teams will compete in making weekly forecasts.

    For this course, taking part in the training phase is sufficient.

    All information for the competition is available from the competition's webpage

    Schedule

    If you are interested in taking part in the challenge, please enroll in the WueCampus course and send an email with all members in your group to Anna Krause. If you don't have a group yet, please use the discussion forum in the WueCampus course.

    Kickoff meeting will be on the TBD. We will discuss possible approaches for the challenge and discuss the schedule for the remainder of the semester. We will also go over available hardware for training models.

    At the end of the semester, each group should be able to present a functioning forecasting model, which is evaluated using the evaluation setting of AI Weather Quest.

     

    Examination

    At the end of the semester, all students present their work in a 20 minute talk with 10 minutes for questions. In addition, a report of 10-15 pages must be submitted.