Downscaling AI weather models to local areas

Type of project

M.Sc.

Contact

Jesper Mariegaard (jem@dhigroup.com)

Topic

In the past couple of years many AI meteorological models have emerged, many of these have similar accuracy as physics-based models but are around 100 times faster. Several of these models are open-source, e.g., Microsoft’s Aurora https://github.com/microsoft/aurora and could probably be used by DHI to develop our own downscaled local weather forecasts (or hindcasts) as input to our water models. There are already several different approaches and it does require some deep learning knowledge to develop and run these models. We imagine a student from ML/AI/data science/computer science/maths could carry out a project with us testing out e.g. two different AI downscaling approaches.

Relevance to DHI

Having high quality meteorological input is relevant for most water modelling done in DHI as it is often the main driver (precipitation and temperature etc for inland models; wind and air-pressure for marine models). With AI models being a 100 times or more faster we could effort running high-resolution local models and potentially in ensemble

What is the student's contribution

Testing out e.g. two different AI downscaling approaches on 1-2 cases. Expected outcome is code and feasibility assessment