Get Started
Home
Topics
Search
Library
Research questionHow can cloud-cover forecasts extend beyond nowcasting despite imperfect NWP initialization while preserving satellite spatial detail?Satellite observations provide a useful initial cloud state, but cloud fields form, dissipate, and deform, causing short-range forecast skill to decline after the first few hours. Longer horizons must represent atmospheric evolution without losing the observed spatial structure when NWP initialization differs from the satellite state.
AI
Computer Vision
Diffusion Models
Evaluation & Benchmarks
Image & Video Processing
Machine Learning
Research Paper
Statistical Machine Learning
Technology
Latest papersRecent research connected to this question, newest first.From Nowcasting to Forecasting: Adapting a Reanalysis-TrainedThe source describes CloudCast v2, which forecasts 12 hours of cloud cover from satellite-derived initial fields and NWP inputs. It reports evaluation against CloudCast v1 over the 1–12-hour range using mean absolute error and fractions skill score, with training on reanalysis data followed by adaptation to satellite cloud fields through conditional flow matching.research paper · Sep 3, 2026
Related questions
How can we improve weather forecasts two to six weeks ahead despite rapidly degrading skill?How can AI weather models produce high-resolution global forecasts without relying solely on analysis data?How can probabilistic weather downscaling recover site-specific conditions from coarse grids despite unresolved terrain and land-surface effects?How can multimodal models improve tropical-cyclone track and intensity forecasts across weather foundation models without retraining?