Data-Driven Modeling and Analysis of Numerical Weather Predictions
MCML Authors
Kevin Höhlein
Dr.
* Former Member
Abstract
Kevin Höhlein
Dr.
* Former Member
Abstract
Weather prediction systems generate vast numerical simulation datasets that require statistical postprocessing and interactive human exploration. In this thesis, we develop deep-learning-based methods for postprocessing weather predictions and representing the forecasts for subsequent analysis. We use neural networks to enhance the spatial resolution of weather forecasts and postprocess ensemble predictions, and adapt neural networks as compact representations for volumetric ensemble datasets.
BibTeXKey: Hoe25