Lecture
SAILS Lunch Time Seminar: Serban Vadineanu
- Date
- Monday 28 September 2026
- Time
- Location
- Online only
Developing a new high-resolution probabilistic weather prediction model using AI
Abstract:
Machine learning is rapidly changing the way weather forecasts are produced. AI weather prediction (AIWP) models learn the evolution of the atmosphere directly from reanalysis data and, once trained, run faster and more cheaply than numerical weather prediction (NWP). This talk presents KNMI's work on probabilistic AIWP.
Our model uses a stretched-grid approach: the grid is refined over the region of interest and left coarser elsewhere, so that a single model produces high-resolution regional forecasts within a global domain, advancing the atmospheric state in six-hour steps. Members are generated via noise injection to the model’s parameters, and the model is trained with the continuous ranked probability score.
Verification against observations for July 2026 shows that the ensemble improves on the HARMONIE-AROME ensemble for 2 m temperature and 10 m wind speed and is competitive with ECMWF's AIFS-ENS, while AIFS-ENS retains an advantage for mean sea level pressure and precipitation at longer lead times.
Join us!
The SAILS Lunch Time Seminar is an online event, but it is not publicly accessible in real-time. Please click the the link below to register to our mailinglist and receive participation links for our Lunch Time Seminars.
Sign up