3B.3 ECMWF Forecast Improvements and Future Perspectives

Monday, 29 January 2024: 2:15 PM
302/303 (The Baltimore Convention Center)
tony McNally, ECMWF, reading, United kingdom

The most recent ECMWF upgrade (Cycle 48R1) was a double landmark in the development of our operational forecasting. Firstly, the system is entirely hosted at our new super-computing data centre in Bologna, and secondly, the full ensemble (ENS) is now run at the same horizontal TCo1279 resolution (approximately 9km grid) as the HRES deterministic forecast. At the same time, this ambitious cycle introduced a wealth of other advances, including a major reformulation of the extended- range system, which now runs with 100 members every day. Focus now turns to the preparation of Cycle 49R1, which has particular significance as it will be used as the scientific basis of our next Copernicus atmospheric climate reanalysis (ERA6) as well as corresponding components for the ocean and atmospheric composition.

Beyond operational upgrades, ECMWF continues to make significant advances towards our longer-term strategic goal of ever higher resolution forecasting and data assimilation, and consistent coupling across all components of the Earth system. This involves a balance between solving the science challenges of operating at the km-scale, but also technical investment in our systems to make them agile and ready to exploit developments in the rapidly evolving High Performance Computing market. In both aspects we collaborate extremely productively in the framework of the European Commission Destination Earth programme.

Finally, the last 12 months has seen an astonishing rise in the successful use of Machine Learning for weather forecasting. It is clear that this disruptive technology presents exciting opportunities to extend and accelerate what we do, but it also represents a direct (and potentially existential) challenge to traditional NWP approaches. These developments have energised ECMWF to investigate radical new directions for our next generation prediction systems.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner