Data-driven Weather Prediction

PI Name Gianmarco Mengaldo
PI Institution Other
Collaborating ANL Division Mathematics and Computer Science (MCS)
Project Description

We aim to test the performance of novel neural network architectures on the WeatherBench dataset and evaluate how they compare against existing methods. We then want to apply the framework to data equal to WeatherBench but at higher resolution, to assess the feasibility of data-driven weather prediction for short and mid-range.

Testbed

Nvidia A100, AMD GPU, CPL “Cooper Lake”, Witherspoon, GPU_V100_SMX2