{"id":"arxiv:2508.21664","kind":"preprint","title":"Trajectory learning for ensemble forecasts via the continuous ranked probability score: a Lorenz '96 case study","authors":["Sagy Ephrati","James Woodfield"],"author_count":2,"year":2025,"published":"2025-08-29","venue":null,"cited_by":null,"doi":null,"arxiv_id":"2508.21664","nct_id":null,"source_url":"https://arxiv.org/abs/2508.21664","archive_url":"https://space-kitty.com/papers/arxiv:2508.21664","updated":"2025-10-22","pmid":null,"pmcid":null,"openalex_id":null,"pdf_url":"https://arxiv.org/pdf/2508.21664","abstract":"This paper demonstrates the feasibility of trajectory learning for ensemble forecasts by employing the continuous ranked probability score (CRPS) as a loss function. Using the two-scale Lorenz '96 system as a case study, we develop and train both additive and multiplicative stochastic parametrizations to generate ensemble predictions. Results indicate that CRPS-based trajectory learning produces parametrizations that are both accurate and sharp. The resulting parametrizations are straightforward to calibrate and outperform derivative-fitting-based parametrizations in short-term forecasts. This approach is particularly promising for data assimilation applications due to its accuracy over short lead times.","abstract_license":"CC0-1.0 (arXiv descriptive metadata)","license":null,"is_open_access":false,"retracted":false,"peer_reviewed":false,"categories":["math.NA","cs.LG"],"details":null,"provenance":"arxiv","attribution":"Abstract: arXiv descriptive metadata (CC0). Thank you to arXiv for use of its open access interoperability. This service was not reviewed or approved by, nor does it necessarily express or reflect the policies or opinions of, arXiv.","topics":[],"cited_in_dossiers":[]}