Physics-Aware Consistency Evaluator (PACE)
Physics-Aware Consistency Evaluator (PACE) is a tool to check how realistic ML-based weather forecasts really are. We test whether models like GraphCast (for global forecasts) and CorrDiff (for local high-resolution downscaling) follow the rules of physics, making sure things like wind and temperature behave consistently over time and across space.
Follow the developments on Github.
Mentors
- Michael Langguth
- Erik Pavel
- Sindhu Vasireddy
- Ankit Patnala
- Savvas Melidonis
- Christopher Goddard
Participants
Martin Vozár
Peter Váš
Marek Rodný
Martin Petrovič