Skip to content

Probabilistic Downscaling of Climate Variables Using Denoising Diffusion Probabilistic Models

License

Notifications You must be signed in to change notification settings

davitpapikyan/Probabilistic-Downscaling-of-Climate-Variables

Repository files navigation

Probabilistic Downscaling of Climate Variables

Project with Colloquium (MA8114) at TUM: Probabilistic Downscaling of Climate Variables Using Denoising Diffusion Probabilistic Models

Supervisor: Prof. Dr. Rüdiger Westermann (Chair of Computer Graphics and Visualization)
Advisor: Kevin Höhlein (Chair of Computer Graphics and Visualization)


Downscaling combines methods that are used to infer high-resolution information from low-resolution climate variables. We approach this problem as an image super-resolution task and employ Denoising Diffusion Probabilistic Model to generate finer-scale variables conditioned on coarse-scale information. Experiments are conducted on WeatherBench dataset by analysing temperature at 2 m height above the surface variable. See the final report here.


References

Releases

No releases published

Packages

No packages published

Languages