Open access
Sep 2026
Deep Learning for CAPE Bias Correction in the NOAA Global Forecast System
A two-step regression–diffusion model (NVIDIA CorrDiff) is applied to address the issue of persistent low-CAPE bias in the operational GFS model, and the effectiveness of CorrDiff highlights its potential for other challenging applications involving small-scale phenomena with long-tailed distributions.
Wei Li, Lin-Lin Cui, Jun Wang et al.
· Meteorology · 0 citations