Overview
- Post-season reporting says the DeepMind model delivered unusually accurate forecasts for both hurricane tracks and intensity in its first operational run.
- Experts highlight that the neural-network approach produces forecasts far faster than physics-based systems that rely on large supercomputers.
- The early comparison did not include ECMWF’s widely regarded physics-based model, so results are considered provisional pending broader benchmarking.
- Veteran forecasters report the U.S. GFS performed poorly this season, with a 2019 dynamic-core upgrade now viewed by some as a setback.
- Some observers have speculated that data-collection lapses tied to government cuts or disruptions could have hurt GFS performance, a link that remains unconfirmed.