In order to bring together real IT problems of existing companies with the fresh ideas of PhD students, ETH and the University of Zurich organise the Academia Industry Modeling (AIM) week once a year. Like its other industrial partners, Meteomatics has given the students a real problem, namely the temporal interpolation of wind fields.
Nowadays we rely on model data, such as the global model from ECMWF or GFS, in order to understand and predict atmospheric processes better. However, the temporal resolution of most model outputs lies between 1 hour and 3 hours and even increases with forecast horizon, although they are run in a much higher resolution. Hence, we do not know what happens in between these time steps, what poses a big uncertainty for, e.g., the energy market or insurances. The energy market is, for instance, interested in 15 minutes steps in order to act accordingly to fast changes. To achieve this, we apply a temporal interpolation between the different time steps for each grid point. However, this does not represent the reality very well and errors often occur because of the interpolation. That is why we posed the challenge to improve the temporal interpolation of wind fields.
The participants came up with new ideas within a limited amount of time. Meteomatics gladly examines this new approaches further and is looking forwards to the next AIM.