High-Resolution Weather Forecasting with StormCast on AMD Instinct GPU Accelerators

Numerical weather prediction is a computationally demanding enterprise, especially at the so-called meso-scales, where convective processes are important.

Machine learning based weather prediction models offer an interesting new possibility to sidestep some of the problems faced by the traditional physics-based models. It is in this respect that NVIDIA’s StormCast, a convection-emulating machine learning based weather prediction model showcases some of the recent rapid progress in the field.

In this blog post we have briefly outlined the problem of weather prediction at different spatial and temporal scales. We discussed the problem of convection and thunderstorm prediction at meso-scales, and what difficulties this poses for traditional methods. We then presented StormCast, explained the basics of how the model works and how it partially solves some of these problems.

Read More: High-Resolution Weather Forecasting with StormCast on AMD Instinct GPU Accelerators — ROCm Blogs